Morning Edition · Saturday, July 18, 2026Published at 1:46 AM EDT · New York
Anthropic Redeploys Fable 5 With a Cross-Lab Jailbreak Severity Scale
The proposed Cyber Jailbreak Severity scale, running from CJS-0 to CJS-4 and developed with Amazon, Microsoft, Google and other Glasswing partners, scores cyber jailbreaks on capability gain, breadth, ease of weaponization and discoverability.
Anthropic brought Claude Fable 5 back online on July 1 after a suspension, following a jailbreak that Amazon researchers found could induce the model to reveal software vulnerabilities. The redeployment included updated cybersecurity safegu…
Continue the AI Intelligence Brief
Track frontier labs, chips, export controls, model releases, regulation, and AI infrastructure.
- 5 AI intelligence signals a day
- Frontier labs, compute, and chips
- Model releases and AI infrastructure
- Source-grounded analysis with confidence labels
The Global Intelligence Brief stays free.
Part of a tracked trend
Oversight and Evaluation Lag Accelerating AI Capabilities
Over the next 3-6 months, evidence mounts that governance, evaluation, and agent-safety methods are failing to keep pace with capability growth, driving investment in interpretability, agent-manipulation benchmarks, and institutional-reform proposals.
More from this edition
- EU Orders Google to Share Search Data With AI Rivals and Open Android Assistant Slots
- NVIDIA Reframes Vera Rubin Around Cost Per Token for the Post-Training Era
- Li Auto's 35B Mixture-of-Experts Model Claims 100B-Class Performance From Post-Training Alone
- Meta's Brain2Qwerty v2 Decodes Typed Sentences From Non-Invasive Brain Signals at 61% Word Accuracy
- Hyundai Moves to Full Ownership of Boston Dynamics Days After Its Workers Struck Over Robots
- Meta Puts Muse Spark 1.1 Behind a Paid API in a Direct Bid for Coding Agents
- Anthropic Restructures Claude Code Review Into Five Effort Tiers, Topped by a Cloud Agent Fleet
- Tongyi Lab's Wan-Dancer Generates Minute-Scale 720p Dance Video From Music
- Kaiser Nurses Say AI Tools and Workplace Surveillance Are Degrading Patient Care
- ReactBench Targets the Gap Between Coding Agents and Real Front-End Work
- University of Chicago Law School Splits Its Curriculum Around AI Rather Than Banning It