In six weeks at the start of 2026, OpenAI began testing ads in ChatGPT, announced the largest private funding round on record and agreed to put its models on the Pentagon’s classified network. In the same weeks a scholar reported that the word “safely” had gone from its mission statement. Each was reported on its own.
You can keep the app. The aim here is that what you use, and what you own through a fund, is a choice rather than a default.
ChatGPT is the thing you asked about your weird dream. It is also made by a company that started in 2015 as a nonprofit research lab and now runs its business through a for-profit corporation. The prompts still work the same way.
what happened on friday night
Late on Friday 27 February 2026, US time, OpenAI chief executive Sam Altman said in a post on X that his company had agreed to deploy its AI models on the classified network of the US Department of Defense. OpenAI was already a Pentagon supplier: in June 2025 the department awarded it a prototype contract with a $200 million ceiling.
The post came hours after President Trump directed every federal agency to stop using Anthropic, the first AI lab on that classified network, and Defence Secretary Pete Hegseth said he was designating it a “Supply-Chain Risk to National Security”. Courts have since split on that label: a San Francisco judge ruled one designation unlawful in August 2026, and a Washington appeals court upheld the other in September.
Anthropic had wanted assurance that its models would not be used for fully autonomous weapons or mass surveillance of Americans. Axios reported that the Pentagon had spent days contending those red lines were philosophical and “woke”. Altman told staff that OpenAI shared the same red lines, and it announced its own deal within hours.
That weekend Altman told followers on X the deal was “definitely rushed” and that “the optics don’t look good”, and Anthropic’s Claude overtook ChatGPT on Apple’s App Store. He said OpenAI “really wanted to de-escalate things”. On Monday 2 March he went further: OpenAI “shouldn’t have rushed”, and the deal “looked opportunistic and sloppy”.
hold on to this one
what a clearance actually buys
Access to classified work. The contract OpenAI already held covered prototypes. The terms agreed that Friday let its models run on the Pentagon’s classified networks. The full contract has not been published, so its full terms cannot be read from outside.
the timeline nobody asked for
Read one at a time, each of these was a news cycle. From the first to the Pentagon post is six weeks to the day.
ads announced
OpenAI says it will start testing ads in ChatGPT within weeks, for free and Go tier users in the US.
the ad test starts
The ad test begins in the US, with ad personalisation switched on by default. It starts the day after Anthropic’s Super Bowl ads made fun of ads in chatbots.
a word goes missing
A Tufts University scholar reports that “safely” has gone from the mission statement in OpenAI’s tax filing for 2024, released in November 2025.
the largest private raise on record
OpenAI announces $110 billion of new investment at a $730 billion pre-money valuation. The round closes on 31 March at $122 billion, valuing OpenAI at $852 billion.
a rival is blacklisted
President Trump directs every federal agency to stop using Anthropic, and Defence Secretary Pete Hegseth says he is designating the company a supply chain risk.
the clearance
Late that night, Sam Altman posts on X that OpenAI has agreed to deploy its models on the Pentagon’s classified network. US strikes on Iran begin hours later.
the optics
TechCrunch reports Altman’s admission that the deal was “definitely rushed” and “the optics don’t look good”. Claude had overtaken ChatGPT on Apple’s App Store the day before.
the rewrite
Altman says OpenAI “shouldn’t have rushed”. The agreement gains a sentence ruling out intentional domestic surveillance of US persons.
what the deal actually says
OpenAI published its account of the agreement on 28 February 2026. It lists three red lines: its technology is not to be used for mass domestic surveillance, to direct autonomous weapons systems, or for high-stakes automated decisions such as “social credit” systems. The deployment is cloud-only, with no models on edge devices. OpenAI says it keeps “full discretion” over its safety stack, with cleared safety and alignment researchers in the loop.
The agreement builds on the 2025 contract, under which the Defense Department said OpenAI would “develop prototype frontier AI capabilities to address critical national security challenges in both warfighting and enterprise domains”. OpenAI described that work as helping with “administrative operations”. The Intercept reported in September 2026, from documents released in a freedom of information lawsuit, that the version signed on 27 February extends that deal, worth up to $200 million, to the military’s classified networks.
OpenAI says the agreement has “more guardrails than any previous agreement for classified AI deployments”, and it said publicly that Anthropic should not be designated a supply chain risk.
The wording OpenAI published lets the department use its system “for all lawful purposes”, with the red lines written as references to existing law and Pentagon policy. Jessica Tillipman, associate dean for government procurement law studies at George Washington University’s law school, wrote that it “does not give OpenAI an Anthropic-style, free-standing right to prohibit otherwise-lawful government use”. OpenAI says the contract fixes those laws as they stand today, and that it could end the contract if the government broke the terms.
After critics said the first wording left room to surveil Americans, OpenAI added a sentence on 2 March 2026: “the AI system shall not be intentionally used for domestic surveillance of U.S. persons and nationals”. NBC News reported the next day that the full contract had not been made public.
the ads in your conversations
Eighteen days before the Pentagon post, on 9 February 2026, OpenAI began testing ads in ChatGPT in the US. They appear below the answer for free and Go tier users, matched to the topic of the conversation. Ad personalisation was switched on by default when the test began, The Register reported. OpenAI says it does not share conversations with advertisers and that ads do not influence answers.
Before the launch, The Information reported that OpenAI was considering prices as high as $60 for every thousand views, a figure OpenAI did not confirm. The test has since left the US: OpenAI’s release notes date the rollout to free and Go users in Australia to 16 April 2026.
Altman wrote on X on 16 January 2026 that “a lot of people want to use a lot of AI and don’t want to pay”. Six weeks later the same company agreed to put its models on a classified military network.
the mission statement that keeps changing
Fortune went through OpenAI’s tax filings and counted six rewrites of its mission statement in nine years. The first, covering 2016 and 2017, opened: “OpenAI’s goal is to advance digital intelligence in the way that is most likely to benefit humanity as a whole, unconstrained by a need to generate financial return.” The latest, for 2024: “OpenAI’s mission is to ensure that artificial general intelligence benefits all of humanity.”
The promise to “openly share our plans and capabilities” was the first to go, in the filing for 2018. In the filing for 2024, released in November 2025, “safely” and “unconstrained by a need to generate financial return” both went. OpenAI said it planned to achieve the mission “primarily by attempting to build safe AGI and share the benefits with the world”.
In January 2024 OpenAI removed “military and warfare” from the uses its policies banned, while keeping a ban on using its tools to “develop or use weapons”.
In February 2026 it disbanded its mission alignment team, six or seven people set up in 2024 to explain the mission to staff and the public. OpenAI called it “a support function” and said the work “continues throughout the organization”. What stands today is a for-profit public benefit corporation, about a quarter owned by the nonprofit OpenAI Foundation.
the water and the silence
OpenAI has not published a sustainability report. As of August 2026 it had not disclosed any of its Scope 1, 2 and 3Three categories of carbon emissions. Scope 1 is what a company produces directly, like its own factories. Scope 2 is from the energy it buys. Scope 3 is everything else in the supply chain. Scope 3 is usually the biggest and the hardest to measure. emissions or made a net-zero pledge, according to a Bloomberg report republished by Fortune. A spokesperson said OpenAI was preparing for a California law that requires large companies to report emissions from November 2026.
The nearest thing to a company figure came in a blog post on 10 June 2025, where Sam Altman wrote that the average ChatGPT query uses about 0.34 watt-hours of electricity and “about 0.000085 gallons of water; roughly one fifteenth of a teaspoon”. A study led by University of California, Riverside researchers, first released in 2023, estimated that GPT-3, an earlier OpenAI model, consumed a 500 ml bottle of water for roughly every 10 to 50 medium-length responses, counting the water used to generate its electricity.
For AI as a whole, a study in the journal Patterns in December 2025 estimated 32.6 to 79.7 million tonnes of CO₂ in 2025, a footprint its author compares to New York City’s. Microsoft, one of the companies OpenAI rents computing from, reported total emissions of 20.29 million tonnes of CO₂ equivalent for the year to June 2025, up 25% on the year before.
the circular money
The funding announced on 27 February 2026 came with supply deals attached. Amazon agreed to invest up to $50 billion, $35 billion of it conditional, and the same day OpenAI agreed to expand its cloud contract with Amazon Web Services by $100 billion over eight years. Nvidia invested $30 billion, and OpenAI committed to 3 gigawatts of dedicated inference capacity and 2 gigawatts of training capacity on Nvidia’s Vera Rubin systems. The investors are also the suppliers. That is neither unlawful nor unusual in technology, but it means part of OpenAI’s valuation rests on spending promised to the companies that set it.
what this means
ChatGPT is still the app you used this morning. You asked it about your weird dream. On the free plan in Australia, it may have shown you an ad afterwards.
The company behind it has changed. A nonprofit research lab founded in 2015 now does its business through a for-profit corporation valued at $852 billion on 31 March 2026, with a classified military agreement, ads beside free users’ conversations and no published emissions figures. Its chief executive announced the Pentagon deal on a Friday night and said within days that it had been rushed.
You cannot yet buy OpenAI shares on a stock exchange. The company said in June 2026 that it had filed confidentially for a listing, and at the end of September Altman said he had no particular timeline in mind. You can already own the companies around it. Microsoft, Amazon and Nvidia are in the S&P 500, which means they are in most global index funds, which means they are probably in your default super fund as well.
inaam holds itself to the same reading. Every company in the fund is named in the app, with the reason it got in and the trade-off written next to it. The three checks a company has to pass are set out on how it works, and the five themes members customise for themselves are on investment areas. The subscription is $10 a month, which is $120 a year, and it is separate from the money you invest. The minimum to invest in the fund is $100. Sign-up happens in the app, not on this website. SMSFs and family trusts are welcome.
OpenAI has not yet published its emissions. The documents linked here are the nearest thing to an account of them.
key sources
OpenAI on its agreement · Axios on the red lines · TechCrunch follow-up · Fortune on the mission · the Patterns study on AI emissions
General information only. It does not take your objectives or circumstances into account. Consider the PDS and TMD, and we recommend independent advice, before investing. This article is general information only. It does not constitute financial advice, a recommendation to buy, sell or hold any financial product, or an endorsement of any investment strategy. Past performance is not indicative of future results. Every claim here is sourced and linked, and where a figure is an estimate or a range, the method and the source are named. Readers should do their own research before making investment decisions. The inaam Impact Investments Fund is ARSN 691 614 132. inaam is not affiliated with OpenAI, Microsoft, Anthropic, Amazon, Nvidia, SoftBank or any entity discussed in this report.
