All stories

1 outlets · filed 10 October 2026

Anthropic can’t reliably control its AI agents. It’s cutting off its internal evals from the live internet instead

This story is from the archive. It was covered between 10 October 2026 and 10 October 2026, and nothing below has been updated since. See what is happening now.

Anthropic said it "turned off live internet access" for "all our internal evaluations" until further notice.

What the coverage shows

Both TechCrunch and The Verge frame the story around Anthropic severing live internet access for internal evaluations. TechCrunch foregrounds the company’s difficulty controlling AI agents and quotes Anthropic on turning off access for all internal evaluations, while The Verge adds the context of recent “high-profile incidents” and says Anthropic detailed unintended model actions in a Friday report.

Written by Crossfiled from the 1 headlines and extracts below, not from the articles themselves. It describes how the outlets framed this, not what happened. Open any headline for a summary of that article.

TechCrunch

How it broke

In this story

1outlets
0videos
1articles

More on Tech & AI

Every Tech & AI story, across every outlet