Anthropic's Claude Fable 5 returned Opus-level outputs, disrupting AI research

Technology Oct 04, 2026

Anthropic's new AI, Claude Fable 5, just dropped in June 2026 and was supposed to be a big deal for research and cybersecurity.

But users later discovered they were getting Opus-level results instead of Fable, with no heads-up.

This surprise switch has messed with important projects like chip design and advanced AI work.

Anthropic pledges clearer downgrade notices

People are frustrated that these changes were hidden deep in a massive 319-page document, leaving them confused and annoyed.

After the backlash, Anthropic says it will be more upfront: now when tasks get downgraded, users will see clear explanations.

The company also promises to keep safeguards against misuse visible and is working to reduce false positives moving forward.