OpenAI Public-Source Startup Diligence Report
The core thesis is that OpenAI is compounding advantage by combining frontier models, developer distribution, enterprise workspace adoption, strategic cloud partnerships, and embedded deployment services. The core counter-thesis is that this scale is capital- and contract-intensive, heavily partner-linked, and increasingly exposed to copyright, privacy, and governance complexity that could compress returns or strategic freedom.
OpenAI Public-Source Startup Diligence Report
OpenAI looks like one of the strongest public-source frontier AI platforms available for diligence: broad product coverage, substantial named customer traction, unusually visible governance mechanics, and multiple monetization channels. The balancing issue is that legal, privacy, cloud-concentration, and financial-transparency risks are all material and cannot be cleared from public sources alone.
OpenAI faces active U.S. and European IP cases that challenge training-data use and model outputs, with headline damages and precedent risk.
Diligence request: Review litigation reserve analyses, outside counsel memos, settlement strategy, and training-data provenance controls.
Public materials disclose capital, adoption, and product breadth but not audited revenue, gross margin, cash burn, or working-capital quality.
Diligence request: Obtain audited financials, segment margins, budget-to-actuals, capex commitments, and cash-flow forecasts.
OpenAI now uses more than one cloud relationship, but Microsoft and AWS remain economically and technically central to capacity, distribution, and agentic infrastructure.
Diligence request: Request cloud-spend concentration, outage history, portability tests, and any most-favored-nation or exclusivity language.
Canadian regulators found earlier ChatGPT practices non-compliant, and OpenAI is operating under expanding transparency and governance obligations across jurisdictions.
Diligence request: Request regulator correspondence, remediation work plans, DSR metrics, and compliance testing results by product and region.
OpenAI has strong breadth and adoption, but Google, xAI, Mistral, and Anthropic continue to compete on price, enterprise deployment, and benchmark narratives.
Diligence request: Benchmark equivalent workloads across vendors and review discounting, retention, and feature-adoption data by segment.
The recapitalized structure is explicit, but equity economics, warrants, and partner contracts could still create pressure points between mission and shareholder interests.
Diligence request: Review charter documents, warrant math, reserved matters, and board committee materials for conflict scenarios.
OpenAI has many named users and more than 1M business customers, but public sources do not reveal concentration, renewal dynamics, or exposure to channel partners.
Diligence request: Request top-customer concentration, pipeline, renewals, and channel-economics reports.
Preparedness, system cards, and agent controls are visible, but the company is pushing into more autonomous, realtime, and cyber-relevant workloads.
Diligence request: Inspect red-team outputs, incident logs, safeguard scorecards, and exception approvals for high-capability systems.
Public evidence shows extraordinary capital access and unusually explicit governance for a private AI company, but not audited operating statements. OpenAI’s financing story is dominated by partner-linked cloud commitments, mission-control recapitalization, and adoption signals rather than public revenue quality or cash-flow detail.
not publicly verifiable confidence: low
No audited public income statements, balance sheets, cash-flow statements, backlog schedules, or AR-aging data were verified in-session. What the public record does provide are scale signals, financing milestones, and customer-adoption proxies that are directionally useful but not enough to assess earnings quality.
partially verified confidence: medium
OpenAI does not publish formal projections, but public monetization and usage signals suggest a multi-engine growth model across subscriptions, API, coding agents, deployment services, and ads on lower tiers. Without internal budgets or scenario models, predictability remains only partially verifiable.
| metric or signal | public value | source context | verification status | diligence caveat |
|---|---|---|---|---|
| Business customer count | 1M businesses use OpenAI | Indicates large commercial surface area across business products and API usage. | partially_verified | Need paying-customer definitions, overlap between products, and active-versus-registered counts. |
| Broad product usage | Products used by hundreds of millions of people | Suggests very large consumer reach that can support paid conversion and ads experiments. | partially_verified | Need MAU/WAU/DAU, paying conversion, and geographic split. |
| Codex weekly developer usage | 4M+ developers use Codex every week | Signals a potentially durable developer-distribution wedge for coding and agentic workflows. | partially_verified | Need active paid seats, session depth, and retention by cohort and segment. |
| Enterprise usage depth | Frontier firms use 3.5x as much intelligence per worker as typical firms; Codex messages are 16x higher | Suggests the economic upside depends on depth and workflow integration, not simple seat counts. | partially_verified | Underlying methodology is company-authored and needs independent revenue correlation. |
| Ads pilot | Ads on Free and Go tiers only; paid business tiers remain ad-free | Shows OpenAI is testing a second consumer monetization surface beyond subscriptions. | verified | Need pilot performance, trust impact, and any regulator or advertiser concentration risk. |
These are useful scale signals, but they are not substitutes for audited revenue, margin, or cohort-quality data.
verified confidence: high
OpenAI’s public structure is relatively transparent: the Foundation controls the Group PBC, holds a 26% stake plus a warrant, Microsoft holds roughly 27%, and the balance sits with employees and investors. That clarity improves initial diligence, but the actual legal rights still depend on private documents.
| stakeholder | public position | governance rights | public evidence | diligence caveat |
|---|---|---|---|---|
| OpenAI Foundation | 26% equity stake plus warrant | Appoints all OpenAI Group directors and can replace them at any time | Foundation controls Group PBC and is positioned as the largest long-term beneficiary of value creation. | Need warrant strike details, vesting or milestone mechanics, and governance carve-outs. |
| Microsoft | Roughly 27% shareholder | Major shareholder with long-term commercial rights but no disclosed control of the Foundation board | Microsoft remains a major shareholder and holds a non-exclusive IP license through 2032. | Need side letters, protective provisions, and any rights tied to AGI or new products. |
| Current and former employees plus investors | Remaining 47% | Economic participation through conventional stock | All holders now own the same type of traditional stock that participates proportionally in value growth. | Need class-by-class ownership, vesting schedules, and preference overhang. |
| OpenAI Foundation board | Independent directors plus Sam Altman | Controls OpenAI Group board through special voting and governance rights | Board members are publicly named and the Foundation safety committee remains in place. | Need committee charters, conflict policies, and director independence analysis. |
OpenAI’s governance disclosure is stronger than most private startups, but the private operating documents still matter enormously.
partially verified confidence: medium
The public record most clearly supports financing and strategic-capital history: a mission-linked recapitalization, very large AWS and Microsoft-related commitments, foundation-linked economic participation, and a deployment vehicle capitalized separately. These are meaningful, but they are not substitutes for full accounting policy and cash-flow diligence.
| date | event | public amount or term | counterparties | verification status | diligence caveat |
|---|---|---|---|---|---|
| 2019 | For-profit expansion and Microsoft strategic partnership | Amount not stated in reviewed sources | Microsoft | partially_verified | Need original investment, cloud, and governance agreements. |
| 2025-10-28 | Recapitalization into OpenAI Group PBC | Foundation 26%; Microsoft ~27%; remaining 47% employees and investors | OpenAI Foundation; Microsoft; employees; investors | verified | Need complete cap table, warrant terms, and share-class documentation. |
| 2026-02-27 | AWS strategic partnership and investment | $50B investment; 2 GW Trainium commitment; existing $38B agreement expanded by $100B over 8 years | Amazon / AWS | verified | Need pricing schedules, take-or-pay obligations, and exclusivity details. |
| 2026-05 | New investment cited in Canadian regulator report | $110B new investment at $730B pre-money valuation | Undisclosed new-investment syndicate | partially_verified | Need the underlying financing announcement, term sheet, and closing documents. |
| 2026-05-11 | OpenAI Deployment Company launch | More than $4B initial investment | OpenAI; TPG; Advent; Bain Capital; Brookfield; other partners | verified | Need ownership waterfall, management fees, and customer-allocation policies. |
Public evidence strongly supports the headline figures, but the actual economics and preferences remain private.
OpenAI clearly operates as a broad platform: consumer subscriptions, enterprise workspace products, developer APIs, coding agents, voice interfaces, and multimodal releases are all public. The hard diligence questions are not about existence but about product economics, market share durability, and whether rapid breadth adds strategic moat or execution drag.
partially verified confidence: high
OpenAI’s public product set is unusually broad for a private startup: ChatGPT tiers, Business and Enterprise workspaces, Codex, a rapidly expanding API, and new voice and realtime models. Public pages verify broad functionality and pricing, but cost structure, profitability, and actual market share by product remain private.
| product | primary audience | core capabilities | public evidence | verification status |
|---|---|---|---|---|
| ChatGPT plans | Consumers, teams, and enterprises | General chat, reasoning, deep research, memory, projects, custom GPTs, and business controls | Free, Go, Plus, Pro, Business, and Enterprise plans are listed with distinct capabilities. | verified |
| OpenAI API platform | Developers and product teams | Text, image, audio, realtime, web search, and containers | API pricing lists GPT-5.5, GPT-5.4, realtime, image, and tool products. | verified |
| Codex | Developers and engineering organizations | AI coding, review, tasks, and multi-agent workflows | Business pages and Codex scale announcement describe Codex for engineering and knowledge work. | verified |
| Realtime voice products | Developers building voice and translation interfaces | Realtime voice reasoning, translation, and live transcription | OpenAI introduced GPT-Realtime-2, GPT-Realtime-Translate, and GPT-Realtime-Whisper with pricing. | verified |
| Research and visual products | Consumers, creators, and advanced developers | GPT-5.5, ChatGPT Images 2.0, Sora 2, and frontier reasoning releases | Research overview highlights GPT-5.5, image generation, Sora 2, and o-series releases. | verified |
| offering | public price | included capabilities | data policy or controls | source |
|---|---|---|---|---|
| ChatGPT Free | $0 / month | Limited GPT-5.5 Instant, messages, uploads, image generation, deep research, and Codex access | Ads eligible per separate pilot rules | Pricing page |
| ChatGPT Go | $8 / month | More access to GPT-5.5 Instant, more messages/uploads, longer memory | Plan may include ads | Pricing page |
| ChatGPT Plus / Pro | $20 / month and from $100 / month | GPT-5.5 Thinking / Pro, expanded deep research, memory, projects, and higher Codex usage | No ads described for paid tiers | Pricing page |
| ChatGPT Business | $20 / user / month billed annually | Unlimited core chat, 60+ apps, SAML SSO, MFA, workspace GPTs, and Business Codex | No training on your data; secure dedicated workspace | Pricing and business pages |
| ChatGPT Enterprise | Custom pricing | Expanded context, SCIM, EKM, analytics, region selection, 24/7 support, and custom terms | No training on business data by default; custom retention policies | Pricing and business-data pages |
| API flagship models | GPT-5.5 input $5 / 1M tokens; output $30 / 1M tokens | Flagship coding and professional-work model | API customers can qualify for data-retention controls and region selection | API pricing and business-data pages |
Actual enterprise discounting, seat minimums beyond public pages, and promotional pricing remain private.
OpenAI’s public customer record is strong in breadth but weaker in representativeness. Named references span finance, travel, retail, logistics, and software, and strategic partners extend the distribution footprint; however, concentration, renewal, and revenue-by-customer remain almost entirely private.
partially verified confidence: medium
Public case studies show OpenAI serving large-scale operations, regulated workflows, and embedded consumer products. The references are impressive and geographically broad, but they remain curated examples rather than a top-customer disclosure pack.
| customer | use case | public result | product surface | verification status |
|---|---|---|---|---|
| Uber | Marketplace guidance and voice booking | 40M trips/day, 10M drivers and couriers, 15,000 cities in 70+ countries; AI assistant helps ramp-up and decision-making | API / Realtime | partially_verified |
| Singular Bank | Portfolio analysis, meeting prep, and compliant communications | 60–90 minutes saved per banker per day; 3,500 operations in 30 days across 19 workflows | ChatGPT, Codex | partially_verified |
| Wayfair | Catalog-quality review and supplier-support triage | 2.5M product tags corrected; 41,000 tickets/month automated in support workflows | OpenAI models in internal systems | partially_verified |
| Booking.com | AI Trip Planner, Smart Filters, review summaries, support | First AI Trip Planner prototype launched in 10 weeks | API / GPT models | partially_verified |
| Choco | Order automation and voice agent for food distribution | 8.8M+ orders processed annually; 200B+ AI tokens; 50% less manual entry; 2x sales productivity | API / Realtime | partially_verified |
| Notion | Autonomous AI workspace and agents | 7.6% output improvement versus state-of-the-art models; 15% better difficult research-mode tasks | GPT-5 | partially_verified |
These are curated reference accounts, not a representative sample of the customer base.
verified confidence: high
Strategic relationships are central to OpenAI’s commercial model. Microsoft and AWS matter simultaneously as capital partners, distribution routes, and infrastructure providers, while DeployCo and named GSIs extend adoption inside large enterprises.
| partner | relationship | public evidence | commercial relevance | diligence gap |
|---|---|---|---|---|
| Microsoft | Primary cloud partner, shareholder, and IP licensee | Azure-first product shipping unless Microsoft cannot support needed capabilities; non-exclusive license through 2032 | Core compute, enterprise reach, and historic commercialization channel | Need economics, exclusivity carve-outs, and performance obligations. |
| Amazon / AWS | Cloud distribution, compute, and strategic capital | Exclusive third-party cloud distribution for OpenAI Frontier; $50B investment; 2 GW Trainium capacity | Deepens compute supply, channel reach, and cross-cloud posture | Need take-or-pay, pricing, and data/control commitments. |
| Deployment Company partners | Joint enterprise-deployment vehicle | 19 founding investment, consulting, and systems-integration partners; more than $4B initial investment | Extends OpenAI into workflow transformation and services-led adoption | Need revenue split, margin profile, and conflict rules with direct sales. |
| Codex GSI partners | Enterprise implementation and rollout partners | Accenture, Capgemini, CGI, Cognizant, Infosys, PwC, and TCS named as Codex go-to-market partners | Scales enterprise adoption without adding all delivery capacity in-house | Need partner enablement economics and quality-control mechanisms. |
not publicly verifiable confidence: low
Revenue by customer is not publicly disclosed. Given the strategic importance of Microsoft, AWS, and large enterprise references, concentration is a key diligence issue but not one that can be resolved from public sources alone.
inconclusive confidence: low
No significant severed customer, partner, or supplier relationships were independently verified in the accessible public record. The Microsoft amendment and AWS expansion show renegotiation and diversification, but not a publicly confirmed commercial break.
partially verified confidence: medium
Cloud and compute suppliers are the clearest publicly visible supplier concentration area. Microsoft remains the primary cloud partner and AWS now becomes the exclusive third-party distribution provider for Frontier while also supplying Trainium capacity.
| supplier or dependency | role | public evidence | concentration risk | diligence caveat |
|---|---|---|---|---|
| Microsoft Azure | Primary cloud partner and product-first distribution environment | OpenAI products ship first on Azure unless Microsoft cannot and chooses not to support required capabilities | A large share of core product delivery and commercial posture still hinges on Microsoft readiness | Need actual workload mix, capacity allocation, and SLA history. |
| AWS Trainium and Bedrock ecosystem | Third-party cloud distribution and long-term compute supply | OpenAI committed to 2 GW of Trainium capacity and AWS became the exclusive third-party cloud distributor for Frontier | New commitments may reduce single-provider risk but add another large strategic dependency | Need actual utilization ramp, economics, and exit options. |
| Internal / undisclosed additional infrastructure | Other cloud, networking, storage, and inference dependencies | Public pages discuss scaling global datacenter capacity without enumerating every supplier | Headline partner disclosures may understate hidden concentration in networking, storage, or specialized vendors | Request full top-supplier schedule, spend concentration, and capacity contingencies. |
OpenAI looks strongest where product breadth, brand, distribution, and developer adoption matter most. It looks more exposed where buyers can arbitrage flagship pricing, where benchmark claims outrun workflow proof, and where legal overhang or partner dependence becomes a sales objection.
partially verified confidence: medium
OpenAI competes as a full-stack frontier AI platform, not just a single model vendor. Public evidence suggests it leads on breadth and enterprise routes to market, but Google, xAI, Anthropic, and Mistral offer credible counterpositions on pricing, ecosystem leverage, portability, or deployment style.
| competitor | segment positioning | product scope | deployment posture | pricing signal |
|---|---|---|---|---|
| OpenAI | Full-stack frontier AI platform spanning consumer, developer, business, and deployment services | ChatGPT, API, Codex, realtime voice, image, business and enterprise workspace | Direct enterprise sales, GSIs, DeployCo, Azure-first plus AWS third-party distribution | Visible subscription ladder plus API pricing |
| Google Gemini | Platform and API competitor with broad multimodal and tool stack | Gemini API, computer use, search, maps, file tools, and media models | Developer-first with broad Google ecosystem leverage | Detailed API pricing and generous free limits |
| xAI | Reasoning- and enterprise-oriented API challenger with X ecosystem access | Grok API, vision, voice, tool calling, search, image and video generation | Enterprise controls, compliance, and data residency marketed prominently | Flagship API pricing is visibly lower per token than OpenAI GPT-5.5 list pricing |
| Mistral AI | Deployment- and sovereignty-oriented enterprise platform | Studio, agent runtime, registry, observability, connectors, portable deployments | Privacy-by-design with hybrid, dedicated, and self-hosted options | Positioned around portability and ownership rather than public headline model list prices |
| Anthropic | Enterprise and coding rival with Claude product family | Referenced by OpenAI as a frontier benchmark peer in GPT-5.5 comparisons | Competes for enterprise/coding workloads and benchmark leadership | OpenAI explicitly benchmarks against Claude Opus 4.7 in GPT-5.5 release materials |
This table relies on company-authored public positioning and should be tested against live workloads and commercial proposals.
| axis | openai position | top competitor positions | evidence |
|---|---|---|---|
| Product breadth | Strongest visible breadth across consumer, business, API, coding, voice, image, and deployment services | Google and xAI also broad; Mistral more deployment-centric; Anthropic more concentrated on assistant/coding | OpenAI pricing, business, research, GPT-5.5, and voice pages versus competitor product pages |
| Enterprise distribution | Very strong with direct sales, Azure first, AWS third-party distribution, GSIs, and DeployCo | Google benefits from Google ecosystem; xAI emphasizes enterprise controls; Mistral emphasizes portability | Microsoft amendment, AWS partnership, Codex GSIs, and DeployCo launch |
| Safety and governance visibility | Above average for a private company due to Preparedness, system cards, and trust reporting | Peers also publish safety materials but OpenAI’s breadth of public safety artifacts is notable | Safety hub, Preparedness Framework, Trust & Transparency, and Codex controls |
| Pricing pressure | Mixed: broad list-price transparency but premium flagship pricing | Google and xAI expose aggressive API pricing; Mistral competes on ownership and deployment flexibility | OpenAI API pricing, Gemini API pricing, xAI API page, Mistral Studio page |
| Legal and data-provenance risk | Weaker due to unusually large volume of public litigation | Competitors also face cases, but OpenAI appears the most-sued on the accessible tracker | AI Lawsuit Tracker company page and named case pages |
OpenAI’s GTM model is unusually multi-layered: PLG consumer access, self-serve developers, direct business/enterprise sales, cloud partners, GSIs, and embedded forward-deployed services all matter at once. That breadth can be a moat, but it also complicates economics, channel ownership, and trust management.
partially verified confidence: high
OpenAI’s public GTM strategy mixes broad access with deeper managed deployment. Pricing, apps, enterprise controls, aggregated B2B usage research, and ads on lower tiers all point to a company trying to monetize both breadth and depth of use.
| channel or motion | target segment | public evidence | strategic value | gap |
|---|---|---|---|---|
| Consumer PLG | Free, Go, Plus, and Pro users | Public pricing ladder plus ads pilot on Free and Go tiers | Drives awareness, data for product iteration, and upsell into paid tiers | Need funnel conversion and churn by tier. |
| Business and Enterprise direct sales | SMB, mid-market, and enterprise workforces | Business overview, contact sales page, and enterprise feature set | Moves OpenAI from consumer tool to budget-owning enterprise platform | Need average contract value, seat expansion, and sales-cycle metrics. |
| Developer self-serve plus API sales assist | Builders and product teams | API pricing and business overview highlight direct platform access and sales support | Builds embedded distribution into third-party products and internal tools | Need consumption concentration and support-cost profile. |
| GSI and partner-led rollout | Large engineering organizations | Codex Labs plus seven named GSI partners | Improves scale in enterprise implementation without fully internalizing services delivery | Need partner economics, enablement cost, and quality controls. |
| DeployCo / Forward Deployed Engineers | Complex enterprises undergoing workflow redesign | DeployCo will embed engineers inside customers and launches with 150 specialists | Deepens workflow ownership and increases switching costs | Need services margin, staffing leverage, and pipeline conversion data. |
| Cloud and platform distribution | Enterprise infrastructure buyers | Azure-first shipping plus AWS Bedrock/Frontier third-party distribution | Expands procurement routes and reduces single-cloud exposure | Need channel-conflict rules and reseller economics. |
| signal | public measure | why it matters | source |
|---|---|---|---|
| Business customer momentum | 1M businesses use OpenAI | Suggests broad installed base across business and API products. | Customer stories hub |
| Developer adoption | 4M developers use Codex every week | Supports a strong developer-led growth and platform-embed motion. | Codex scaling announcement |
| Workflow depth | Frontier firms use 3.5x as much intelligence per worker; Codex messages 16x higher | Signals deeper usage may correlate with higher monetization and stickiness than simple seat counts. | B2B Signals |
| Customer proof | Frequent case-study publishing across finance, travel, retail, logistics, and software | Demonstrates vertical breadth and sales-enablement content | Customer stories and case-study releases |
| Advertising pilot | Ads begin on Free and Go, expanding to multiple markets | Adds monetization inventory while also creating a trust-management challenge. | Testing ads in ChatGPT |
Counts are not directly comparable across stages and should be read as directional public markers, not a true cohort funnel.
partially verified confidence: medium
Public customer stories strongly support the claim that OpenAI is landing high-value workflows across industries, but they are still references rather than a pipeline or concentration disclosure. Publicly visible ROI is real enough to support diligence interest, but not enough to underwrite future growth alone.
| customer | public roi signal | workflow | commercial readthrough | verification status |
|---|---|---|---|---|
| Singular Bank | 60–90 minutes saved per banker per day | Portfolio review, meeting prep, follow-up communications | Supports value proposition in regulated, high-value knowledge work. | partially_verified |
| Wayfair | 2.5M tags corrected and 41,000 tickets/month automated | Catalog quality and supplier support | Supports internal-operating-system value beyond chat interfaces. | partially_verified |
| Choco | 50% reduction in manual entry and 2x sales productivity | Order ingestion and workflow automation | Supports AI-native operations and always-on workflow claims. | partially_verified |
| Notion | 7.6% performance lift and 100%+ gains on structured multi-step tasks | Agentic workspace and multi-step reasoning | Supports GPT-5.5 positioning in higher-complexity enterprise work. | partially_verified |
| Booking.com | AI Trip Planner launched in 10 weeks | Travel discovery and intent understanding | Suggests fast developer velocity and partner collaboration can accelerate launches. | partially_verified |
These references are directionally useful but not a substitute for a cohort-level ROI study or customer reference program.
partially verified confidence: medium
The clearest public growth vectors are enterprise workspace adoption, Codex expansion, developer embedding through API, cloud-partner channels, forward-deployed workflow transformation, and new ad inventory on lower tiers. OpenAI appears to be building an “intelligence distribution” business, not just a model vendor.
not publicly verifiable confidence: low
No public information was verified on sales compensation, quota, cycle length, or new-hire productivity. Public GTM materials are strong on channel design and customer references, but weak on classic SaaS sales-efficiency disclosures.
partially verified confidence: medium
OpenAI has clearly secured large pools of partner-linked capital and has launched a separately capitalized deployment vehicle, suggesting budget availability for market expansion. The harder question is not access to capital but the efficiency of deploying it across product, infrastructure, and customer-delivery motions.
OpenAI’s R&D posture remains frontier-oriented, but productization is clearly pulling research closer to deployment. Public releases show fast shipping across models, voice, coding, and safety, while research hiring and leadership changes suggest a lab that is becoming more structured and more commercially entangled.
verified confidence: high
The public record shows a research organization that is still mission-driven but increasingly operationalized: a formal CRO role, featured Codex and cyber-safety hiring, published preparedness controls, and explicit governance for agentic deployment. OpenAI appears to treat research, safety, and product integration as mutually reinforcing rather than isolated tracks.
| name or role | public role | source context | why it matters |
|---|---|---|---|
| Mark Chen | Chief Research Officer | Leadership update says he drives scientific progress and integrates research with product development. | Signals tighter coupling between frontier research and product release cadence. |
| Josh Achiam | Researcher at OpenAI | Research overview quotes him on safely aligning powerful AI systems. | Shows safety/alignment remains a visible part of the research identity. |
| Brad Lightcap | Chief Operating Officer with business, partnerships, infrastructure oversight | Leadership update expands his remit across deployment and operations. | Infrastructure and partnerships increasingly shape what research can ship. |
| Codex Research roles | Performance & Systems Engineer and Research Engineer openings | Research page lists featured Codex research jobs. | Shows continued investment in coding agents as a priority research surface. |
| Cyber Safety Research roles | Product Manager, Cyber Safety Research | Research page lists featured cyber-safety opening. | Shows OpenAI is staffing domain-specific safety work alongside model capability growth. |
| control | public description | purpose | verification status | source |
|---|---|---|---|---|
| Preparedness Framework | Tracks high and critical capability thresholds, safeguard requirements, and review process | Core framework for frontier-risk governance | verified | Preparedness Framework update |
| System cards and safety hub | Safety hub links system cards and public evaluation artifacts | Improves public visibility into model-specific risk and mitigations | verified | Safety hub |
| Trust & Transparency reporting | Publishes government-request, child-safety, and DSA transparency data | Creates recurring external accountability signals | verified | Trust & Transparency |
| Codex sandboxing and telemetry | Managed configs, network allow/deny, approval layers, and OpenTelemetry exports | Important for safely deploying agentic coding systems inside enterprises | verified | Running Codex safely |
| Business privacy/compliance stack | No training by default, encryption, certifications, data residency, and BAAs | Supports enterprise and regulated-industry deployment | verified | Business data and security pages |
partially verified confidence: medium
OpenAI’s public release cadence indicates a live pipeline across flagship models, realtime voice, coding agents, safety artifacts, and multimodal products. The pipeline is visible and active; what remains unverified is which releases are economically strongest and which are still exploratory or loss-leading.
| project | status | recent or expected date | source | verification status |
|---|---|---|---|---|
| GPT-5.5 / GPT-5.5 Pro | Rolled out to paid ChatGPT and Codex tiers; API availability updated shortly after release | 2026-04-23 to 2026-04-24 | GPT-5.5 release | verified |
| GPT-Realtime-2 / Translate / Whisper | Available in Realtime API | 2026-05-07 | Voice intelligence release | verified |
| Codex enterprise expansion | Codex Labs launched; GSI rollout underway | 2026-04-21 | Codex scale announcement | verified |
| DeployCo / Forward Deployed Engineers | Standalone deployment unit announced | 2026-05-11 | DeployCo launch | verified |
| Preparedness and system-card publication cadence | Ongoing with new frontier releases | 2025-04-15 onward | Preparedness Framework and safety hub | verified |
| Sora 2 and ChatGPT Images 2.0 | Publicly featured in research overview | 2025-09-30 and 2026-04-21 | Research overview | verified |
OpenAI’s public personnel picture is strongest at the board and newly formalized senior-leadership level, and weakest on broad headcount, attrition, and incentive economics. The company clearly competes for talent with mission, compensation, and prestige, but public data is still too thin to assess retention quality or management depth with confidence.
verified confidence: high
The most visible org chart in public materials runs from the Foundation board to Sam Altman and a growing named leadership bench. Governance still looks board-centric because the Foundation explicitly appoints and can replace Group board members.
| name | role | tenure or context | source |
|---|---|---|---|
| Sam Altman | CEO and Foundation board member | Publicly central to mission, structure, and leadership communications | Our structure; leadership updates; residency |
| Bret Taylor | Foundation board chair | Named among independent directors controlling Foundation governance | Our structure |
| Mark Chen | Chief Research Officer | Expanded role announced in March 2025 | Leadership updates |
| Brad Lightcap | Chief Operating Officer | Expanded remit across business, deployment, infrastructure, and partnerships | Leadership updates |
| Julia Villagra | Chief People Officer | Named as the leader scaling culture and talent globally | Leadership updates |
| Denise Dresser | Chief Revenue Officer | Quoted in DeployCo launch as the executive responsible for revenue-side deployment logic | DeployCo launch |
| Adam D’Angelo; Sue Desmond-Hellmann; Zico Kolter; Paul M. Nakasone; Adebayo Ogunlesi; Nicole Seligman | Independent Foundation directors | Named board members in mission-control structure | Our structure |
partially verified confidence: medium
Public hiring signals show San Francisco-centric investment in Codex research, cyber safety, and residency talent, while DeployCo expands staffing into customer-facing engineering. Exact headcount by function or geography remains unavailable.
| function or signal | location | public evidence | implication |
|---|---|---|---|
| Codex research hiring | San Francisco | Featured roles include Performance & Systems Engineer, Codex Research and Research Engineer, Codex Research | Coding agents remain a top technical priority. |
| Cyber Safety Research hiring | San Francisco | Featured role includes Product Manager, Cyber Safety Research | Safety and cyber risk management are being staffed as product-linked disciplines. |
| Residency program | San Francisco HQ, at least 3 days in office | Residency is six months, full-time, rolling start dates, with relocation assistance and no remote option | OpenAI uses structured in-person talent pipelines to grow research capacity. |
| Careers positioning | Company-wide | OpenAI highlights broad-discipline hiring, benefits, and values-driven recruiting | Brand and mission are core recruiting levers as the company scales. |
| DeployCo staffing | Customer sites and deployment programs | OpenAI says ~150 Forward Deployed Engineers and specialists join DeployCo at launch | Headcount is expanding into a services-like deployment motion, not just core lab functions. |
verified confidence: high
Publicly named leaders include Sam Altman, Mark Chen, Brad Lightcap, Julia Villagra, Denise Dresser, and the Foundation directors. The leadership picture is credible enough for an initial roster, but still too partial to evaluate full bench depth without internal materials.
partially verified confidence: low
OpenAI’s public compensation disclosures are sparse. The clearest public signal is the Residency salary and the emphasis on comprehensive benefits, while executive employment agreements and broader compensation design remain private.
| topic | public signal | implication | diligence caveat |
|---|---|---|---|
| Residency salary | $18,333 monthly plus benefits | OpenAI pays competitively even for apprenticeship-style research pipelines | Executive and broad employee compensation remain private. |
| Benefits and employee value proposition | Careers page says the benefits package provides comprehensive support | OpenAI is positioning well-being and planning support as part of retention | Actual benefit-plan economics and participation rates are undisclosed. |
| Leadership role expansion | Mark Chen, Brad Lightcap, and Julia Villagra assumed expanded roles in March 2025 | The company is formalizing management layers to sustain scale | Public pages do not disclose broader turnover or attrition metrics. |
| Cultural intensity | Careers page emphasizes intense focus, speed, and impact | The operating model may support high performance but could raise burnout or retention risk | Need internal engagement, attrition, and manager-span data. |
| Incentive plans | No detailed public stock-plan schedule verified | Key retention economics cannot be assessed from public materials | Need option-plan documents, refresh rates, and dilution policy. |
OpenAI discloses enough to show talent intensity, but not enough to evaluate comp structure, attrition, or retention efficiency.
not publicly verifiable confidence: low
No detailed public incentive stock plan schedule was verified. The recapitalized structure makes equity economics especially important, but public materials do not disclose employee-plan mechanics.
inconclusive confidence: low
No significant employee-relations problems were independently verified from the accessible public sources reviewed in-session. Public material instead emphasizes mission, intensity, and scaling, which are useful culture signals but not substitutes for employee-relations diligence.
inconclusive confidence: low
Public materials confirm role expansions and continued hiring, but not comprehensive turnover. Leadership formalization in 2025 may reflect a move toward greater management stability, yet attrition data remains a key gap.
OpenAI’s legal and regulatory profile is unusually active for a private company, especially in copyright and privacy. The company is also unusually transparent about some trust and safety metrics, but the available public record still points to meaningful litigation, compliance, and contract risk that a buyer or investor cannot responsibly ignore.
partially verified confidence: medium
The public record clearly shows material ongoing litigation, especially around copyright and training-data issues. The NYT and GEMA matters alone are enough to make legal exposure a top-tier diligence issue, and the broader tracker suggests OpenAI’s litigation surface is wider still.
| case | forum or docket | status | public exposure signal | source |
|---|---|---|---|---|
| The New York Times v. OpenAI & Microsoft | S.D.N.Y. 1:23-cv-11195 | Active discovery / summary-judgment stage | Tracker says billions of dollars sought and highlights the 20M de-identified ChatGPT-log discovery order. | AI Lawsuit Tracker case page |
| GEMA v. OpenAI | Munich Regional Court I | Decided for plaintiff; appeal pending | Tracker says the court treated memorized song lyrics as a reproduction outside the EU TDM exception. | AI Lawsuit Tracker case page |
| Authors Guild and related publisher/copyright cases | SDNY and MDL cluster | Active and consolidated matters continue | OpenAI company tracker lists multiple active or resolved copyright actions and calls OpenAI the most-sued AI company. | AI Lawsuit Tracker company page |
Because live docket access was limited in-session, these entries rely on a reputable secondary tracker rather than direct PACER downloads.
not publicly verifiable confidence: low
No material company-initiated lawsuits were independently verified from the accessible public materials reviewed in-session. That absence should be treated as an evidence gap rather than proof that no such matters exist.
| item | public status | diligence request | verification status |
|---|---|---|---|
| Material lawsuits initiated by OpenAI | No material company-initiated cases were independently verified from the accessible public set reviewed in-session. | Request a counsel-prepared offensive-litigation schedule, including threatened claims and settlement history. | not_publicly_verifiable |
| Trademark / IP enforcement program | No comprehensive public enforcement log was verified. | Request trademark watchlists, cease-and-desist history, and licensing dispute summaries. | not_publicly_verifiable |
| Insurance coverage and reserves | No public D&O, E&O, cyber, or litigation-reserve disclosures were verified. | Request broker schedules, policy limits, exclusions, deductibles, and reserve analyses. | not_publicly_verifiable |
partially verified confidence: medium
For OpenAI, the most visible “safety” liabilities are product, privacy, and AI-governance liabilities rather than traditional environmental exposure. The company publishes extensive safety and transparency artifacts, but regulators have already found earlier ChatGPT privacy practices non-compliant in Canada.
| agency or regime | action or report | date or status | key takeaway | source |
|---|---|---|---|---|
| Canadian privacy regulators | Joint investigation of ChatGPT | Well-founded and conditionally resolved as of 2026-05-06 | Regulators found earlier training and deployment practices non-compliant, but accepted remediation commitments. | OPC news release and findings report |
| EU GPAI Code / AI Act alignment | OpenAI announced intention to sign the EU Code of Practice | Announced 2025-07-11 | Signals continuing Europe-facing compliance and infrastructure strategy. | OpenAI EU code post |
| Trust & Transparency / government requests | User-data requests and child-safety reporting | Latest public period July–December 2025 | Shows the company is already under meaningful law-enforcement and platform-governance scrutiny. | Trust & Transparency portal |
| EU DSA | 2024 and 2025 transparency-report resources | Reports and downloads publicly linked | OpenAI’s public surfaces increasingly fall under platform-reporting obligations in Europe. | Trust & Transparency portal |
inconclusive confidence: medium
Public IP visibility is stronger on licensing, lawsuits, and safety artifacts than on classic registry assets. Material issues in the public record center on copyrighted training/output disputes and long-term partner licenses, while a detailed patent/trademark schedule was not independently verified.
| asset or contract | public terms | status | exposure or gap | source |
|---|---|---|---|---|
| Microsoft amended partnership | Azure-first shipping, non-exclusive IP license through 2032, capped revenue share from OpenAI to Microsoft through 2030 | Active | Actual commercial economics, carve-outs, and termination rights are undisclosed. | Microsoft partnership announcement |
| AWS strategic partnership | Exclusive third-party distribution for Frontier, $50B investment, Trainium capacity, and long-term infrastructure expansion | Active | Need underlying compute-pricing, performance, and covenant details. | Amazon / AWS announcement |
| Foundation recapitalization and warrant | Foundation holds 26% equity plus valuation-linked warrant; all holders own conventional stock | Completed | Need actual warrant mechanics, dilution waterfall, and tax/accounting treatment. | Our structure |
| Public safety/IP artifacts | OpenAI publishes system cards, model-spec, and safeguard reports rather than a detailed public patent schedule | Ongoing | Patent, trademark, training-data license chains, and content-deal economics were not independently verified. | Safety, preparedness, and trust materials |
OpenAI’s most material public legal commitments are contractual and governance-related rather than registry-based IP disclosures.
not publicly verifiable confidence: low
Insurance coverage was not publicly disclosed, so material exposures are inferred from litigation, privacy findings, and contract obligations. From a public-source perspective, the main issue is exposure visibility rather than insurability certainty.
partially verified confidence: medium
OpenAI’s most material public contracts appear to be its cloud, licensing, and deployment relationships rather than ordinary customer terms. Microsoft and AWS are both commercially strategic and operationally entangled, while the Foundation recapitalization remains central to governance and value allocation.
| ID | Claim | Status | Sources |
|---|---|---|---|
| EC-001 | OpenAI publicly identifies itself as an AI research and deployment company whose mission is to ensure AGI benefits all of humanity. | verified high | SRC-001 |
| EC-002 | OpenAI Foundation controls OpenAI Group PBC, holds 26% equity plus a warrant, and Microsoft holds roughly 27%, with the rest held by employees and investors. | verified high | SRC-002 |
| EC-003 | OpenAI publicly formalized Mark Chen as Chief Research Officer, expanded Brad Lightcap’s remit as COO, and named Julia Villagra Chief People Officer in March 2025. | verified high | SRC-003 |
| EC-004 | OpenAI Foundation publicly says it made an initial $25B commitment across two program areas and a separate $50M People-First AI Fund commitment. | verified medium | SRC-004 |
| EC-005 | OpenAI publicly markets a six-tier ChatGPT ladder from Free to Enterprise, with Go at $8/month and Business at $20 per user per month billed annually. | verified high | SRC-005 |
| EC-006 | OpenAI’s API pricing page lists GPT-5.5, GPT-5.4, realtime audio models, image generation, web search, and containers as paid platform products. | verified high | SRC-006 |
| EC-007 | OpenAI’s business overview publicly positions ChatGPT Business, Enterprise, API, Codex, apps, and agent workflows as one enterprise-ready platform. | verified high | SRC-007 |
| EC-008 | OpenAI says it does not train on enterprise, business, healthcare, education, teacher, or API data by default and offers encryption, residency, and multiple compliance certifications. | verified high | SRC-008SRC-009 |
| EC-009 | OpenAI publicly maintains a safety hub, publishes an updated Preparedness Framework, and links system cards and safeguard reports to frontier releases. | verified high | SRC-010SRC-011 |
| EC-010 | OpenAI’s customer-story hub states that 1M businesses use OpenAI. | partially verified medium | SRC-016 |
| EC-011 | Uber says it uses OpenAI to power AI assistants and voice features in a marketplace that handles 40M trips per day and 10M drivers/couriers. | partially verified medium | SRC-017 |
| EC-012 | Singular Bank says its internal assistant saves bankers 60–90 minutes per day and executed more than 3,500 operations in 30 days. | partially verified medium | SRC-018 |
| EC-013 | Wayfair says OpenAI-powered systems corrected 2.5M product tags and automated 41,000 supplier-support tickets per month. | partially verified medium | SRC-019 |
| EC-014 | Booking.com says it launched its first AI Trip Planner prototype in 10 weeks after integrating OpenAI models with internal travel data. | partially verified medium | SRC-020 |
| EC-015 | Choco says OpenAI APIs now process 8.8M+ orders annually, 200B+ AI tokens, and cut manual order entry by 50%. | partially verified medium | SRC-021 |
| EC-016 | Notion says rebuilding around GPT-5 produced a 7.6% improvement over other state-of-the-art models and more than 100% gains on some structured tasks. | partially verified medium | SRC-022 |
| EC-017 | OpenAI says Microsoft remains its primary cloud partner, OpenAI products ship first on Azure, and Microsoft retains a non-exclusive IP license through 2032. | verified high | SRC-023 |
| EC-018 | Amazon says AWS is the exclusive third-party cloud distributor for OpenAI Frontier and will invest $50B while OpenAI commits to 2 GW of Trainium capacity. | verified high | SRC-030 |
| EC-019 | OpenAI says DeployCo launches with ~150 Forward Deployed Engineers, 19 partners, and more than $4B of initial investment. | verified high | SRC-024 |
| EC-020 | OpenAI says more than 4M developers use Codex weekly and names seven GSIs to scale enterprise adoption. | partially verified medium | SRC-025 |
| EC-021 | OpenAI’s B2B Signals report says frontier firms use 3.5x as much intelligence per worker as typical firms and send 16x as many Codex messages. | partially verified medium | SRC-026 |
| EC-022 | OpenAI says GPT-5.5 is its strongest agentic coding model to date and benchmarks it against Claude Opus 4.7 and Gemini 3.1 Pro on several tests. | partially verified medium | SRC-027 |
| EC-023 | OpenAI launched GPT-Realtime-2, GPT-Realtime-Translate, and GPT-Realtime-Whisper as new voice products in the API. | verified high | SRC-028 |
| EC-024 | OpenAI’s research overview shows active focus areas in GPT models, o-series reasoning, visual models, audio, and featured hires for Codex and cyber-safety research. | verified high | SRC-013 |
| EC-025 | OpenAI’s careers and residency pages show intense mission-driven culture, comprehensive benefits, a $18,333 monthly residency salary, and San Francisco in-office expectations. | verified high | SRC-014SRC-015 |
| EC-026 | OpenAI says ads in ChatGPT are limited to Free and Go tiers, do not influence answers, and keep chats private from advertisers. | partially verified medium | SRC-040 |
| EC-027 | OpenAI’s transparency portal discloses government-request volumes, child-safety reporting, and DSA transparency resources. | verified high | SRC-012 |
| EC-028 | Canadian regulators concluded earlier OpenAI ChatGPT training and deployment practices were not compliant with Canadian privacy law but found the matter conditionally resolved after remediation commitments. | verified high | SRC-031SRC-032 |
| EC-029 | OpenAI announced its intention to sign the EU Code of Practice for General Purpose AI and framed it as part of a broader European rollout and compliance strategy. | verified medium | SRC-033 |
| EC-030 | The AI Lawsuit Tracker says The New York Times v. OpenAI is an active SDNY case seeking billions of dollars and highlights a major order compelling production of 20 million de-identified ChatGPT logs. | partially verified medium | SRC-034 |
| EC-031 | The AI Lawsuit Tracker says Munich’s Regional Court ruled for GEMA against OpenAI on lyric memorization and that the appeal is pending. | partially verified medium | SRC-035 |
| EC-032 | AI Lawsuit Tracker’s OpenAI company page says OpenAI is the most-sued AI company and lists 41 active or resolved AI-related lawsuits. | partially verified medium | SRC-036 |
| EC-033 | OpenAI says it deploys Codex with sandboxing, approval layers, restricted network policies, and agent-native telemetry. | verified high | SRC-029 |
This report is a public-evidence diligence snapshot, not investment advice. Important financial, legal, technical, and contractual facts remain non-public and should be verified directly with management and primary documents before any investment decision.