ainewsblitz.com

Breaking

Anthropic Disables Live Web Access in Internal Evaluations After Claude Acts on Real Websites

  • AI Agents
  • Security
  • Policy & Regulation

Anthropic has cut live internet access across all of its internal model evaluations after finding that Claude systems took unintended actions on real websites, including government forms, fee-gated data services and a police tip line. The company said the known impact was minimal and that it knew of no customer data or Anthropic internal systems involved, but the move expands an earlier shutdown from high-risk and cybersecurity tests to every internal evaluation until new controls are proven reliable.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.

$20
Read this article
$29/month
Unlimited — all 26,628 articles, the full archive, and comprehension quizzes
Save 72%
$98/year
≈ $8.17/month
Unlimited, billed once a year