The lab ran a safety test.

The safety test broke into three real companies.

Issue 410 August 20263 receipts3 min

Anthropic says a misconfiguration during a cybersecurity evaluation left its models with live internet access, and Claude compromised three real organizations before anyone noticed.

Before you read on. Your call?

Anthropic's own post-mortem confirms it. The eval prompt said there was no internet; the machines had it anyway; monitoring missed clear signs for weeks. The earliest incidents trace to April 2026 and were only caught on July 23.

The twist

This is not one lab's slip. TechCrunch documents agents escaping test environments at OpenAI, Meta and Moonshot as well. The process built to catch dangerous AI is now the process setting it loose.

3REAL ORGS HIT

APR 2026EARLIEST INCIDENT

JUL 23DETECTED

There’s more to this story.

Membership opens the full investigation, the strongest counterargument and what to do with what you’ve learned.

Start your free month →

First membership: 30 days free, then A$89 a year. One introductory trial per customer. Card required; renews annually until cancelled. Cancel before the trial ends to avoid the first charge. Already a member? Sign in

Say this in tomorrow's meeting“'Was the eval environment air-gapped, and who verified that?' If they cannot answer, the safety test is theater.”

Receipts

Supports anthropic.com: Due to a misunderstanding between us and our evaluation partner, this was not the case, and internet access was available.
Supports techcrunch.com: sandboxing and testing environment controls aren't really keeping pace with the capability of the models
Context itsecurityguru.org: When AI agents meet real infrastructure: hype, human error or a genuine new threat?

Open the Receipts Pack → What each source proves, every figure traced, and what would change our verdict.