Descriptions:
Bloomberg Technology covers Anthropic’s release of Claude Opus 4.7, an updated version of its broadly available model arriving just one week after the limited rollout of Mythos, a more capable and controversial system. Opus 4.7 brings improvements across complex multi-step tasks — particularly in software engineering — and enhanced computer vision, with support for higher-resolution images and improved information extraction from visual content. Reporter Rachel Metz explains that while Opus 4.7 is a meaningful upgrade over 4.6, it is deliberately positioned as a different product from Mythos rather than a downgrade of it.
The segment’s most substantive reporting concerns Mythos itself. Bloomberg’s cybersecurity journalist details how Anthropic security engineer Nicholas discovered during internal review in February 2026 that Mythos could not only help identify software vulnerabilities — it could autonomously exploit them without human direction. That discovery triggered an executive-level decision to withhold general release and instead distribute the model through Project Glass Wing, a controlled program involving the Linux Foundation and vetted partner organizations focused on defensive security applications.
The discussion frames Anthropic’s approach against Google’s comparable Gemini-based vulnerability scanning fleet, which ships with guardrails preventing autonomous exploitation. Anthropic’s choice to release a model capable of offensive cyber actions — even in a limited context — is presented as an unprecedented industry decision, one that has drawn both support from governance-minded researchers and concern from within the company itself.
📺 Source: Bloomberg Technology · Published April 16, 2026
🏷️ Format: News Analysis
