Anthropic Fights to Bring Most Powerful AI Models Back Online

Anthropic’s latest AI models, released last week, are already taking some forced PTO under pressure from the White House.

Amazon CEO Andy Jassy raised a red flag with Trump administration officials last Thursday, claiming the new models’ guardrails had gaps that hackers could use to find security flaws in digital infrastructure. The Commerce Department issued an export control directive to Anthropic on Friday, forcing the company to shut down Fable 5 and Mythos 5. 

Anthropic, which defended its models’ safety standards, is rushing to resolve the dispute, with senior tech executives reportedly flying to DC yesterday for meetings.

A Lose-Lose Situation

The bad blood between Anthropic executives and the Trump administration is getting thicker. The two parties are coping with a failure to communicate that stems at least in part from core ideological differences. President Trump previously called Anthropic employees “left-wing nutjobs” around the same time the Pentagon labeled the company a supply-chain risk and banned defense contractors from using its products. 

The messy relationship isn’t just a bad look for the US but potentially a big stumble backward:

• Cybersecurity experts say the shutdown of Anthropic’s models could open bigger holes in the US’s digital defenses than the models themselves did. Leaders from companies including Nvidia and Adobe signed a letter Sunday asking Washington to lift the export controls, arguing the models aren’t unique in being able to find security flaws and that coders count on them to defend against potential attacks from rival AI models. 
• Some of those rival models come from China, including Kimi 2.7 and potentially, per the letter, secret Chinese government AI projects with even greater capabilities. The letter says that China’s AI models are just months behind the US’s. The implication: While the US restrains Anthropic, China could use that time to close the gap. 

Optics Matter: US leaders locking horns with homegrown AI companies may raise concerns that Washington will impede AI’s progress, benefiting runner-up China. Shares of Chinese AI model-makers surged on Monday, with the listed entity behind Zhipu climbing as much as 48% and rival MiniMax jumping nearly 8%. While China’s lack of lithography tools is likely to limit its AI advancements, its models could benefit from affordability and reliability. DeepSeek’s AI model costs 60 times less to run than Anthropic’s Fable 5, geoeconomics expert Agathe Demarais pointed out, and if Washington starts shelving American AI models with little notice, consumers might favor models that they can count on when needed.

The post Anthropic Fights to Bring Most Powerful AI Models Back Online appeared first on The Daily Upside.