When OpenAI first announced its AI defense deal with the Pentagon, CEO Sam Altman claimed if all went well, the firm could end up looking like “geniuses” who managed to de-escalate the situation between the US government and the AI industry after a fallout with Anthropic. Anthropic had just backed out from signing a deal that – at least on the surface – looked mostly identical on moral and ethical grounds.
In the same beat, Sam Altman had also cautioned that the deal might look rushed, sloppy and opportunistic on the outside. Either way, OpenAI has found itself in the middle of a big controversy. As one OpenAI employee has himself gone on record to say, the deal was “not worth it”, circumstantial evidence from both inside the company and outside of it, seems to suggest that the ChatGPT-maker may have bitten off more than it could chew.
Let’s start with the inside. The OpenAI scientist, who said the deal wasn’t worth the trouble, was only the tip of the iceberg, though to be fair, he acknowledged that the system at the company allowed employees to voice their opinion and that he was proud of this culture. In an all-hands meeting, Sam Altman reportedly told staff that OpenAI and its employees may not have a lot of say in how the Pentagon used its AI technology following the deal.
His exact words, according to a Bloomberg report, were, “You do not get to make operational decisions.” Anthropic’s demands for having a say in how its AI systems were used by the military (after the deal) may have been the root cause of tensions between the Dario Amodei-led firm and the Pentagon. Anthropic had walked out of the deal because it was concerned that the military might use Claude for mass domestic surveillance and development of autonomous weapons.
Sam Altman admits OpenAI has no direct say
The US has used Anthropic’s AI systems during its strikes on Iran and the capture of Venezuelan President Nicholas Maduro. According to CNBC, Sam Altman addressed such incidents and explained, “So maybe you think the Iran strike was good, and the Venezuela invasion was badYou don’t get to weigh in on that.” In other words, while OpenAI could provide the safety stack and technical guardrails, it could not veto specific legal military operations and say what should and what should not be done.
Do note that the Pentagon and OpenAI have stated that the AI systems will be used only for “all lawful purposes.” Sam Altman has said that the company has amended its rushed contract to ensure its red lines are not crossed.
OpenAI VP joins Anthropic amid mass user exodus
OpenAI’s deal with the Pentagon has sparked outrage online. ChatGPT uninstallations surged by almost 300 per cent as users boycotted the OpenAI chatbot. Reports indicate ChatGPT lost over 1.5 million subscribers in the 48 hours following the deal. Many switched to Anthropic’s Claude chatbot to show support for the company’s stance against the Pentagon.
Turmoil may have begun even within OpenAI. Previously, 100 employees of OpenAI released a joint statement alongside 871 Google employees, supporting Anthropic and protesting the use of AI in military surveillance.
Now, OpenAI’s VP of research, Max Schwarzer has left the company to join Anthropic. While Schwarzer did not blame the deal for his exit, his post on X suggests that he was more inclined towards Anthropic’s “values.” He wrote, “I have also been very impressed with Anthropic’s talent, research taste and values.” It does indicate that perhaps Anthropic’s refusal to accept Pentagon’s demands may have played a role in attracting Max.
Though Sam Altman has reiterated that he has opposed the Pentagon’s decision to label Anthropic a supply chain risk.
OpenAI stops deployment to US defense agencies
Following the widespread backlash online, OpenAI has decided to stop the deployment of its AI systems to US defense agencies such as the National Security Agency (NSA).
OpenAI staff member Noam Brown wrote on X, “OpenAI will not be deploying to the NSA or other DoW intelligence agencies for now.” Brown accepted that the rushed agreement left “questions unanswered” such as “some novel ways that AI could potentially enable legal surveillance.”
According to Brown, the language of the contract has been updated. OpenAI will “address these loopholes through the democratic process” before it deploys its systems for these agencies.
Dario Amodei slams Sam Altman over OpenAI deal
Following Sam Altman’s deal announcement with the Pentagon, Anthropic boss Dario Amodei claimed that OpenAI’s arguments defending the deal were full of lies. In an internal memo, Amodei stated that the public was able to see through the “gaslighting” attempt made by Altman. He hoped that OpenAI employees could see this as well.
Amodei also stated that despite Altman’s efforts to become a peacemaker, the public was seeing Anthropic as “heroes” for refusing to accept the Pentagon’s demands.
