Anthropic’s Dario Amodei accuses Sam Altman of ‘gaslighting’; labels OpenAI’s Pentagon deal as ‘safety theater’: Report

Anthropic CEO Dario Amodei expressed his messages regarding the agreement he made with the Pentagon to strengthen OpenAI’s secret work. Notably, OpenAI announced its agreement with the US government last week, just hours after its rival Anthropic was blacklisted by the US government. Even CEO Sam Altman admitted that the timing of OpenAI’s deal with the Pentagon seemed ‘opportunistic and careless’ and later regretted the timing of the announcement.

Altman also spent the last few days making statements about the Pentagon deal, noting that OpenAI’s deal with the government has more safeguards than the one with Anthropic. But Amodei isn’t buying this explanation, having written a memo to his staff stating that OpenAI’s messages were ‘blatant lies’ and that Altman had falsely “presented himself as a peacemaker and deal maker”.

The memo has not yet been made public and was reported by The Information. Amodei also called OpenAI’s dealings with the Department of Defense “security space.”

“The real reason [OpenAI] accepted [the DoD’s deal] “We didn’t do that because they cared about calming down employees and we actually cared about preventing abuse,” Amodei said.

“I guess this attempt at distortion/gaslighting doesn’t work well with the public or the media, where people mostly view OpenAI’s deal with the DoD as sketchy or questionable and view us as heroes (we’re currently #2 on the App Store!),” Amodei wrote.

“It’s working on some Twitter morons, which is fine, but my real concern is how to make sure it’s not working on OpenAI people,” he added.

Anthropic’s growing popularity:

Notably, shortly after the deal was announced, ChatGPT faced public backlash from users who threatened to terminate their OpenAI subscriptions in favor of Anthropic. In fact, Anthropic’s Claude has dethroned ChatGPT to become the #1 free app in the App Store rankings in several countries.

ChatGPT removals rose 295% day-over-day on Saturday, a day after the Department of Defense deal was announced, according to another TechCrunch report citing data from market intelligence provider Sensor Tower.

Meanwhile, Claude’s downloads increased 37% each day on Friday and 51% on Saturday. Since then, Anthropic has tried to ease the transition to Claude by introducing an import memory feature to bring ChatGPT or other chatbot memories into Claude. The company also continued to make the memory feature available to free users, in a move aimed at attracting OpenAI users.