Anthropic's magic code-sniffer: More Swiss cheese than cheddar, for now

The Register critiques Anthropic's Mythos, an AI code security model, finding it effective at automating detection of known vulnerability classes but unable to identify novel flaws humans would catch. Project Glasswing's limited rollout is presented as responsible, though the article questions whether the tool's impact justifies the accompanying hype. It frames Mythos as one step in an inevitable evolution toward AI-assisted security engineering working alongside human expertise.