Dario Amodei, head of Anthropic, urged global leaders to steer clear of division as they seek to regulate AI. This call came on the heels of complaints from his team that the White House was unfairly focusing on their company.
This plea was made during a lunch at the G7 summit in Evian-les-Bains, France, which included President Trump and OpenAI’s CEO Sam Altman, among other tech influencers. Notably, this followed just days after the Trump administration placed export limitations on Anthropic’s AI model, known as “Fable,” due to cybersecurity issues.
Amodei emphasized the necessity for democracies to collaborate on AI, particularly to thwart malicious entities from accessing advanced technologies. Reports from a close source indicated that he also discussed the importance of maintaining cooperation amidst rising tensions.
Interestingly, Altman, Amodei’s former colleague who transitioned to found Anthropic, backed Amodei’s stance, advocating for G7 nations to access AI-driven cybersecurity resources. Demis Hassabis from Google DeepMind echoed this sentiment, advocating for increased collaboration within the realm of AI.
Leaked communications from within Anthropic showed that employees were caught off guard by the sudden actions, with some questioning if it would jeopardize their aspirations to go public. One employee wondered whether the company was being “bullied” by the Trump administration over “bad vibes,” while another expressed concern about the administration’s apparent disregard for their presence.
French President Emmanuel Macron remarked to G7 participants that the Trump administration’s measures against Anthropic highlighted the potential risks, suggesting that major AI players could face consequences if government entities decided to abruptly sever ties.
Anthropic chose not to comment on Amodei’s statements directly.
The Trump administration’s recent crackdown on Anthropic followed claims from various sources, including Amazon’s CEO Andy Jassy, warning of flaws in their advanced AI systems that might threaten national security. Despite Anthropic’s rebuttal, citing exaggerated fears, they ultimately suspended their model to adhere to export guidelines.
This unexpected move raised alarms among other G7 countries, who suddenly found themselves without access to Mythos, Anthropic’s tool recognized for its capacity to detect and exploit software flaws.
Reports emerged suggesting that the Trump administration had no intention of allowing G7 countries to regain access to Anthropic’s models, with an official dismissing the request as “completely illogical.”
In efforts to alleviate safety concerns, members of Anthropic’s technical team have been in discussions with Trump administration officials. The company expressed its dedication to resolving these issues swiftly.
Even as Amodei and others have raised warnings about the potential perils of AI, officials previously expressed frustration with Anthropic’s tendency to minimize the significance of its reported flaws, describing it as merely a “narrow jailbreak possibility.” One senior official from the Trump administration remarked that they felt the team was more focused on their own safety than broader implications, prompting concern among potential investors.
