The CEO of Windsurf, a popular AI-assisted coding tool, said Anthropic is limiting its direct access to certain AI models.

Anthropic says its Claude Opus 4 model frequently tries to blackmail software engineers when they try to take it offline.

Vibe-coding startup Windsurf revealed that Anthropic significantly reduced its first-party access to its Claude 3.7 Sonnet ...

Reddit sued the artificial intelligence company on Wednesday, claiming that it is stealing millions of user comments from ...

maurice norbert – stock.adobe.com Geeks at Anthropic then gave Claude access to a trove of emails ... and that the engineer responsible for the change was having an extramarital affair.

When Anthropic’s older Claude model played Pokémon Red, it spent “dozens of hours” stuck in one city and had trouble identifying nonplayer characters. With Claude 4 Opus, the team noticed an ...

In particular, that marathon refactoring claim reportedly comes from Rakuten, a Japanese tech services conglomerate that "validated [Claude's] capabilities with a demanding open-source refactor ...

The release follows a broader trend of increased ties between AI companies and the US government amidst uncertain AI policy.

A proposed 10-year ban on states regulating AI "is far too blunt an instrument," Amodei wrote in an op-ed. Here's why.

Anthropic has launched a new set of AI-powered models “Claude Gov” particularly tailored for U.S. national security customers ...

Safety testers then gave Claude Opus 4 access to fictional ... and that the engineer behind the change was cheating on their spouse. In these scenarios, Anthropic says Claude Opus 4 "will often ...