Caitlin Kalinowski resigned over the deliberation and oversight behind OpenAI’s Pentagon agreement. The company says its safeguards prohibit domestic surveillance and autonomous weapons, leaving a dispute over how those limits were established and will be enforced.
Caitlin Kalinowski, OpenAI’s hardware leader and former head of Meta’s augmented-reality glasses team, resigned over the governance of its Pentagon deal.
OpenAI says cloud controls and contractual restrictions protect against surveillance and autonomous weapons; it added surveillance language before her departure.
The agreement extends an existing defense business, whose earlier pilot had a $200 million contract ceiling.
Caitlin Kalinowski announced her resignation on March 7, 2026, saying OpenAI had rushed its Pentagon announcement before defining guardrails. The departure brings an internal leadership objection to the AI developer’s move into classified military environments, even as OpenAI says its agreement protects the boundaries she raised.
Kalinowski joined OpenAI in November 2024 after leading Meta’s augmented-reality glasses team, according to the report confirming her departure. That background explains the significance of losing a hardware leader; the report does not establish that she negotiated the military agreement.
In her resignation announcement, she supported a national-security role for AI but objected to the deliberation given to surveillance of Americans without judicial oversight and lethal autonomy without human authorization. She expressed pride in the robotics team’s work and respect for OpenAI CEO Sam Altman and her colleagues.
In a follow-up post quoted in the report, she said the announcement preceded defined guardrails and described her objection as one of governance. OpenAI confirmed her departure in a statement to the publication, saying the agreement supported responsible national-security uses while prohibiting domestic surveillance and autonomous weapons.
Her departure establishes a disagreement about safeguards and decision-making, rather than evidence that OpenAI technology has been used for either disputed purpose.
Police are investigating a top tax-accountant exam scorer's admission of AI-assisted cheating, while the government reviews national qualification tests held in 2026 and prepares stronger controls.
Editorial Team
The protections OpenAI says it retained
OpenAI’s February 28 explanation says it reached the classified-deployment agreement the previous day. The company describes three boundaries: no mass domestic surveillance, no use of its technology to direct autonomous weapons systems, and no high-stakes automated decisions such as social-credit systems.
Its claimed enforcement mechanism combines contract terms, existing law and technical control. OpenAI says deployment is cloud-only, that it runs and can update the safety controls, and that security-cleared engineers and safety researchers remain involved. It says it will not supply models with safeguards removed or deploy them on edge devices, arguing that those constraints prevent use in fully autonomous weapons.
That is OpenAI’s technical assurance. The published contract extract is more qualified than a blanket weapons prohibition: it bars independent direction of autonomous weapons where law, regulation or department policy requires human control, and other high-stakes decisions requiring human approval under those authorities. OpenAI’s broader assurance therefore rests on the combined operation of its technical controls, contractual language and legal restrictions.
A March 2 update says the parties added language explicitly barring intentional domestic surveillance of U.S. persons and nationals, including deliberate tracking or monitoring through commercially acquired personal or identifiable information. OpenAI also said Pentagon intelligence agencies such as the NSA would need a new agreement to use its services.
OpenAI also said it had initially held back from a classified-deployment contract because it did not consider its safeguards and systems ready. Its account presents the agreement as the result of work to address that problem.
Those additions preceded Kalinowski’s resignation. They show that OpenAI had published surveillance restrictions before she left, while leaving her criticism of the deliberation and timing unresolved.
An incumbent rival and an existing defense business
Anthropic, the developer of the Claude AI models and chatbot, was already a military supplier. In a February 26 statement, the company said it had been the first frontier AI developer to deploy models on U.S. government classified networks. It described Claude’s national-security work as including intelligence analysis, operational planning and cyber operations. Anthropic argued that surveillance law had not caught up with AI’s capabilities, making legal compliance alone an insufficient boundary in its view.
The March 7 account describes failed negotiations over safeguards against mass domestic surveillance and fully autonomous weapons, followed by the Pentagon’s designation of Anthropic as a supply-chain risk. Anthropic said it would challenge the designation in court. Microsoft, Google and Amazon said they would keep Claude available to non-defense customers, limiting the scope of the commercial fallout described in that report.
OpenAI argues that its agreement offers stronger protections than Anthropic’s earlier contract. That comparison is the company’s claim, not an independently established ranking of the two suppliers’ safeguards. OpenAI also says it opposed Anthropic’s supply-chain designation and asked the government to offer the same terms to other laboratories.
OpenAI’s own defense relationship predates this deal. In its June 16, 2025 government initiative announcement, it described a pilot through the Defense Department’s Chief Digital and Artificial Intelligence Office, with a $200 million ceiling. The proposed work covered administrative operations, service-member and family health-care access, acquisition data and proactive cyber defense, subject to OpenAI’s usage policies.
The ceiling measures the potential size of that earlier pilot. It does not establish money received or the value of the classified-environment agreement. The relevant change is expanded access to sensitive military work within an existing business relationship.
OpenAI’s company-reported ceiling for its earlier defense pilot was $200 million; this does not establish money received or the later agreement’s value. Graphic: Magica. Source: Introducing OpenAI for Government.
What would demonstrate enforcement
On recourse, OpenAI says in its explanation of the agreement that it could terminate the contract if the government violated it, and that the agreement incorporates the referenced laws and policies as they stood when agreed, preserving those standards if they later change.
Its March 2 update also describes a planned Pentagon working group involving AI laboratories, cloud providers and policy and operational officials. That offers a forum for discussion, rather than evidence that deployment controls have worked.
The next test is whether the promised controls identify and stop prohibited uses in practice. Evidence about how proposed uses are reviewed, disputed cases are escalated and violations trigger restrictions or termination would help assess that question.
An early tally puts Japanese listed companies’ first-half net profit growth at 43.2%, led by manufacturers. Filings show AI component demand alongside currency benefits and non-AI earnings gains, while procurement delays and canceled infrastructure reservations expose the costs of expansion.
TSMC plans to back promising chip startups through a new investment arm, alongside separate commitments to OpenAI and Anthropic. The strategy reaches toward future customers and the AI developers shaping chip demand, but the investment rights remain unclear.
Nidus has raised $30 million across pre-seed and seed rounds to automate wire processing in existing U.S. defense factories. Its founders bring manufacturing and government-deployment experience, but customer access and proprietary data still need to translate into measurable production gains.
Nebuly is funding US and European expansion for its AI business-value platform. Adjustable assumptions and private deployment are part of its pitch, but rivals offer similar measurement tools and verified customer returns remain undisclosed.
Han Won-seok says he used a miniature camera, generative AI and a smartphone in tax accountant and CPA exams. His confession puts scrutiny on exam security after proctors reported no suspicious behavior.
Thousands protested against proposed AI data centres in Scotland as ministers prepare new planning guidance, putting public participation and the scale of electricity demand at the centre of the dispute.
Uber is notifying drivers after a breach at AI infrastructure provider Fireworks exposed personal and driver’s license information. Three reported state counts add up to about 7,870 people. Fireworks has since disclosed wider data exposure, but the precise connection used to reach Uber’s records remains unexplained.
Jensen Huang’s reported resignation puts Nvidia’s financing and advocacy for American open models under scrutiny. Researcher Nathan Lambert warns of a loss of support, while Nvidia’s commercial incentives and continuing model work complicate that forecast.
Altimeter and Atreides want Nvidia's board to leave Jensen Huang's return open while it investigates recent events. Their letter backs interim leadership and acknowledges that directors can ultimately reject his reinstatement.
Elon Musk says Nvidia risks losing supply arrangements made personally by former CEO Jensen Huang. No affected supplier is named. Nvidia’s filing documents both $95.2 billion in commitments and limits to the protection those arrangements provide.
The former Anthropic and OpenAI researcher says a pretraining-efficiency slide helped drive his resignation, turning an investor selling point into an argument for a development pause.
Anthropic’s IPO has reportedly attracted orders for six times the available shares at proposed terms targeting a roughly $1.4 trillion valuation. The sale would add to substantial existing cash reserves as the Claude developer balances costly capacity shortages against long-term computing commitments.
Jason Clinton is urging network isolation as AI accelerates vulnerability discovery. The PHP emergency illustrates the immediate repair problem, while Anthropic's own security work shows that finding flaws does not automatically reduce risk.
Former Nvidia CEO Jensen Huang portrayed his relationship with a customer’s executive as routine business, according to reviewed messages and unnamed sources. A board inquiry faces questions about his disclosures amid separate concerns over chip exports and alleged spying.
Apple has detailed how Secure Exclaves isolate background camera and microphone processing from the main operating system and verify visible indicators before releasing full sensor data. The design aims to keep raw ambient sensing data beyond the main operating system’s reach, but its formal verification remains incomplete.
Charta Health is expanding its insurer tools with a $55 million Series B. Its promise of billing accuracy across payers and providers faces a harder test: showing that better coding reduces disputes rather than simply changes payments.
Operand has raised $31 million across seed and Series A funding for AI due diligence software aimed at private equity firms. Its pitch is to reconstruct businesses from public records and internet traces, but evidence of accuracy and deal-level economics remains limited.
The startup formerly known as QuilrAI pairs enterprise AI agents with Guardian agents and charges by agents under management. Its expansion pitch faces established alternatives and unanswered questions about security performance and cost.
Luminal’s Standard Capital-led Series A backs software for running AI models more efficiently on existing GPUs and alternative chips. Its reported processing volume leaves revenue and comparable customer savings unresolved.
