By Bruce Schneier
Publication Date: 2026-05-08 12:00:00
LLast month, Anthropic made a notable announcement about its new model, Claude Mythos Preview: It was so good at finding vulnerabilities in software that the company didn’t release it to the public. Instead, only a select group of companies would be able to scan and repair their own software.
The announcement requires context – but it contained an essential truth.
While Anthropic’s model is really good at finding software vulnerabilities, other models are also very good at it. The UK’s AI Security Institute found that OpenAI’s GPT-5.5, which is already generally available, is comparable in terms of performance. The Aisle company reproduced Anthropic’s published results with smaller, cheaper models.
At the same time, Anthropic’s refusal to publicly release its new model makes a virtue of necessity. Mythos is very expensive to operate and the company does not appear to have the resources for a general release. What better way to improve company valuation than a reference to…
