Claude Opus 5.5 (Product): news timeline & CVEs

Anthropic reconfigures its cool kids security program

Anthropic merged Glasswing and its Cyber Verification Program into three tiers that change Claude security refusals.

Anthropic combined Project Glasswing and its Cyber Verification Program into Defense Access, Red Team Access, and Specialized Access, with different Claude refusal rules. It said partners verified at least 129,000 vulnerabilities between April and July 2026, more than 33,000 high or critical, and its own scanning found about 5,500 more through October. Of 5,674 true positives, 3,014 were high and 1,522 critical, yet only 516 were patched. A VulnCheck researcher said fewer than 0.5 percent of 225 Anthropic-linked vulnerabilities were exploited in the wild.

The Register · Securityupdated Oct 7, 20:51 UTC · 1d agofirst Oct 6, 23:29 UTC · 2d agoAI safety & security in the wild 16 sources1

Claude Sonnet 5.5

Anthropic released Claude Sonnet 5.5, over 30% faster and up to 30% cheaper while beating Sonnet 5.

Anthropic released Claude Sonnet 5.5, priced the same as Sonnet 5 but described as running more than 30% faster and costing up to 30% less for most work. Simon Willison reports it appears to beat Sonnet 5 on every benchmark and approaches Opus 5.5 on some coding tasks, including viral 3D animation prompts. A pelican SVG test at max thinking effort hit the same 128,000-token bug as Opus 5.5 and failed after costing $1.28; an xhigh run succeeded in 41 seconds for 5.74 cents.

Simon WillisonSep 28, 22:07 UTC · 10d agoModel release1

Claude Opus 5.5

Timeline2 articles

Appears with

Anthropic reconfigures its cool kids security program

Claude Sonnet 5.5