OpenAI is back-pedalling after criticism that it took an “opportunistic and sloppy” deal from the US Government to allow the military to use their technology with allegedly very limited restrictions.

This comes following praise of Anthropic who refused to engage in the same deal with the US Department of War (DoW), with President Trump then demanding all federal agencies stop any use of Anthropic’s tools in Government environments.

The original agreement between OpenAI and the DoW appeared to allow mass surveillance and AI controlled weapons use (so long as they adhered to legal requirements). 

Anthropic failed to come to an agreement, despite stating they believed in the “existential importance of using AI to defend the United States and other democracies, and to defeat our autocratic adversaries”. 

However, misalignment around mass domestic surveillance and fully autonomous weapons prevented the organisation from reaching an agreement with the DoW. 

Anthropic CEO Dario Amodei said that “in a narrow set of cases, we believe AI can undermine, rather than defend, democratic values”, and it was these concerns that allegedly hindered the AI firm’s opportunity to partner with the Government. 

On the 2nd ofMarch, Sam Altman, CEO of OpenAI said they would be altering the previously agreed deal with the US DoW, making amendments to language in the contract to ensure that they prohibited spying on American citizens. 

Altman suggested that these agreements are “super complex, and demand clear communication”. The rushed arrangement has caused significant backlash for OpenAI, from which it’s now scrambling to course-correct. 

Here is re-post of an internal post:

We have been working with the DoW to make some additions in our agreement to make our principles very clear.

1. We are going to amend our deal to add this language, in addition to everything else:

“• Consistent with applicable laws,…

— Sam Altman (@sama) March 3, 2026

What’s been the effect of the botched deal?

Over the last few days, many Open AI customers have cancelled ChatGPT subscriptions which, according to market intelligence firm Sensor Tower, have increased by 295% day-over-day, with many seemingly making the move to Anthropic’s AI platform, Claude – the free version of the app is currently sitting atop the list of most downloaded in the US.

Despite amendments being made to the agreement, concerns still remain. The contract appears to reference the law as preventing mass surveillance of US citizens, however if the laws were to be changed then this would raise serious concerns about how the technology could be used or the loopholes that could be exploited to extend use cases of the technology by the military. 

Contract wording points to the restriction of “deliberate” and “intentional” use, however some are concerned that the military could claim accidental use should a breach arise. 

Recommended reading

• Anthropic Reaches $380bn as AI Investment Frenzy Rolls on
• Comment | What To Expect if the AI Bubble Bursts
• Nvidia Shatters Forecasts as AI Bubble Debate… Bubbles

Use cases of AI tools by the military are vast, from managing immense amounts of information, creating operation simulations, analysing satellite information, to managing resources and logistics. 

Whilst technology such as data analytics tools have been used in military operations for decades, there are concerns about the rapid increase in powers and capabilities AI tools could create for Governments. 

Further to this, becoming overly reliant on AI in military operations poses significant risks due to AIs tendency to make mistakes and hallucinate. Human oversight is essential, particularly given the life or death scenarios the military are dealing with. Allowing AI to make decisions and act on our behalf without oversight could be fatal.  

Altman has tried to manage the backlash by indicating that “follow-on modifications” will be made to the contract to ensure Government agencies do not have uninhibited access to OpenAI, and that any additional US Government agency who wishes to utilise OpenAIs tools will have to create a new, separate agreement first.

It remains to be seen whether these amendments will be enough to restore user faith and alleviate US citizens’ concerns in how their information is being used by the military.