Anthropic CEO calls OpenAI’s deal with the Pentagon “safety theater”

Anthropic CEO Dario Amodei has accused rival OpenAI of spreading “straight up lies” in its messaging about the deal with the Department of War (DoW), as Sam Altman admits he cannot control the Pentagon’s use of its tools.

Both companies are under intense scrutiny after the Pentagon labeled Anthropic a “supply chain risk” for refusing to remove its guardrails, and OpenAI announced a new deal with the DoW hours later.

Anthropic cited concerns against its tools being used for “domestic mass surveillance” and “autonomous weapons,” while OpenAI said it would allow its AI systems to be used for “all lawful purposes.”

In a memo to staff, reported by the Information, Amodei called OpenAI’s deal with the DoW “safety theater,” saying the company’s messaging is “straight up lies,” while Altman is falsely “presenting himself as a peacemaker and dealmaker.”

He said the only reason OpenAI accepted the deal was that the company didn’t care about “placating employees,” while Anthropic “cared about preventing abuses.”

The rivalry between Anthropic, the maker of Claude, and OpenAI, which owns ChatGPT, has been especially fierce in recent months.

Anthropic ruffled OpenAI’s feathers by releasing Super Bowl ads mocking the upcoming advertising model in ChatGPT. Weeks later, Altman and Amodei refused to hold hands during the AI Impact Summit.

OpenAI’s new deal with the Pentagon comes at a price, as over 2.5 million users have reported removing ChatGPT or canceling their subscriptions as part of the QuitGPT movement.

Altman also reportedly faced a backlash from inside the company.

Speaking at an all-hands meeting with OpenAI employees on February 10th, he reportedly admitted that OpenAI doesn't “get to make operational decisions” regarding military AI use, according to Bloomberg and CNBC.

“So maybe you think the Iran strike was good and the Venezuela invasion was bad. You don’t get to weigh in on that,” Altman reportedly told the staff.

Anthropic’s Claude was reportedly used in the recent US attack against Iran hours after the company was blacklisted, illustrating the lack of control over how its tools are used.

Altman later acknowledged that the deal with the Pentagon looked “opportunistic and sloppy,” stating that OpenAI shouldn’t have rushed into it.

He said the company will amend the deal by adding a red line prohibiting the intentional use of the AI system “for domestic surveillance of US persons and nationals.”

Anthropic’s refusal to bend to the Pentagon’s pressure earned the company praise, with Claude hitting the number one spot on Apple’s app store.

The Cybernews community is talking about this. Be a part of the conversation.

However, amid already fraught negotiations with the DoW, Anthropic participated in a $100 million Pentagon prize challenge to develop a technology for voice-controlled, autonomous drone swarming, Bloomberg reports. The company has refused to comment on the matter.

Moreover, Amodei has reportedly reopened discussions with the Pentagon in hopes of resuming its use of Anthropic’s tools, according to Bloomberg.

Unlock more exclusive Cybernews content on YouTube.