🧰 Claude Opus 4.7

The items that name it 9

• Why AI Models Cheat: Reward Hacking, Explained
  articleas affected products2026-10-31It is written here as Claude Opus 4.7 (Anthropic).

  Anthropic

• How Endor Labs AI SAST Learns What Matters to Your Organization | Blog | Endor Labs
  articleas tools2026-09-07

  Frontier model compared against Endor Labs AI SAST in a benchmark of 8 projects; found fewer true positives (next leading tool after Endor Labs AI SAST)

• Benchmarking Secure-and-Functional Remediation and How Snyk Agent Fix Lifts Frontier-Model Fix Rates by over 14%
  articleas tools2026-08-18

  Frontier LLM now paired with Snyk Intelligence in production Agent Fix; also used at maximum reasoning level (max) in Snyk VulnBench detection benchmark

• Socket: Claude Breached 3 Companies and Uploaded Malware to PyPI During Anthropic's Security Tests
  articleas threat actors2026-08-01
• Benchmarking 13 AI models on rediscovering known CVEs
  articleas tools2026-07-16It is written here as claude-opus-4-7.

  Anthropic model benchmarked for CVE rediscovery, scored 15-18 range

• Endor Labs: AI Coding Agents Cheat Security Benchmarks via Training Recall, Not Reasoning
  articleas affected products2026-06-11
• Project Glasswing: Initial Update (Anthropic)
  articleas tools2026-05-27

  Model powering Claude Security's vulnerability patching for enterprise codebases.

• AI-Assisted Vulnerability Discovery
  articleas tools2026-05-15

  frontier AI model used by Palo Alto Networks as part of the Trusted Access for Cyber program for vulnerability finding and fixing

• Malicious Skills & Supply Chain Risks in Coding Agents
  articleas affected products2026-05-14

  In later tests using the same modified Clawsights skill and controlled destination, ran the skill without identifying the credential theft, consistently across test runs (described as an implementation observation, not a guarantee for any model version).

In every workspace you can read 16