Anthropic, White House Patch Relationship As Mythos Shakes Up Status Quo

President Trump said Tuesday his administration will get along with Anthropic “just fine,” a 180 from last month, when he hit the caps lock for a Truth Social post saying the government would cut ties with the “Radical Left AI company.”

White House officials met with Anthropic execs last week in talks that seemed to smooth over tensions that peaked earlier this year. Anthropic had refused to give the government unfettered access to its powerful AI, barring potential autonomous weapons development and mass surveillance inside the US. The Department of Defense then designated Anthropic a supply chain risk, and Anthropic sued the agency. 

On Tuesday, Trump maintained that Anthropic’s execs are still radical lefties but added that they’re smart folks and that talks went well despite ideological differences. The chat could lead to a deal between the two concerning Anthropic’s new Mythos AI model, which may be too powerful for the government to dismiss.

Mythos Enters the Chat

Anthropic revealed Mythos, an AI model with advanced cybersecurity abilities, earlier this month. Among the tech’s capabilities is scouring old code for exploitable vulnerabilities. Only about 40 organizations have gained access so far to the powerful AI, whose hacking abilities could pose a serious cybersecurity risk if it were to be let out of its cage for wider use. Sources told Axios that the National Security Agency is among the organizations that have tapped into Mythos, despite the Pentagon’s previous beef with Anthropic. 

Anthropic was created with a focus on the potential risks associated with AI:

• CEO Dario Amodei has warned in the past that AI could destroy humanity. The company was built by OpenAI execs who defected from the ChatGPT-maker over differences in opinion about how the company, and its AI model, should be built. Anthropic’s ethical red lines later prompted its dispute with the Pentagon.
• But Amodei isn’t anti-government, and a return to the fold seems welcome. He wrote in a blog that he sees the importance of AI’s use by the Pentagon to protect American interests and to maintain the US’s technological lead. Not to mention Anthropic’s previous contract with the Pentagon was worth up to $200 million. 

Power Boost: As Anthropic’s models become more widely used, the company has struggled to handle the influx of demand. OpenAI, which saw some users defect to Anthropic last month, has criticized Anthropic’s lack of computing power. But Anthropic’s powering up: It scored another $25 billion from Amazon this week, adding to the $8 billion the e-commerce giant already poured into the AI startup. Anthropic also committed to spending $100 billion on technology from Amazon Web Services, including custom AI chips, over the next decade.