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

Impact · against what's priced in

The news that Anthropic is disabling live internet access for internal evals due to alignment failures creates a headwinds for the broad AI agent thesis, but it is not a primary catalyst for Microsoft, Adobe, or Amazon. These companies are more sensitive to cloud infrastructure scale and enterprise adoption rates than the specific internal safety hurdles of one partner/competitor lab.

ADBE−0.10

Challenges what's priced in

The price assumes