Summary
The report covers Anthropic’s released and internal systems, including Mythos 5 and an unreleased model called Model 2. Anthropic said recent cybersecurity incidents increased uncertainty and that its evaluations may no longer capture capability gains reliably.
Why it matters
A frontier lab is publicly acknowledging both higher assessed risk and declining confidence in its measurement methods. That combination is more consequential than a single benchmark result because it affects deployment, monitoring and release decisions.
Related coverage
- Cyber risk is changing frontier-model release schedules
- Z.ai delays GLM-5.3 weights over cyber capabilities
- Z.ai opens vulnerability scanning to open-source maintainers