Anthropic warns GLM-5.3 may spread advanced cyber capabilities

Anthropic published a report warning that GLM-5.3 could spread advanced cyber capabilities, raising concerns about security risks from open models.

• Open Models Debate: Weighing Safety Against Democratization2026-10-13
• Study Finds Speculative Reward Hacking Common in Coding Agents2026-09-29
• Agent harness affects cache efficiency and energy use2026-10-17
• Replit ends Anthropic partnership, CEO cites cost and performance2026-10-16
• Anthropic quietly pitches its own AI safety bill to lawmakers2026-10-16