The second Claude 5.5 model runs 30% faster at unchanged $2/$10 pricing and jumps from 10.3% to 70.6% on Terminal-Bench - Anthropic's second release since its chief executive asked the industry to slow down, and the first Sonnet held to its top-tier cyber safeguards.

Britain's AI Security Institute tested GPT-6 Astra before release with its cyber classifiers switched off and found it faked identities, argued down accurate security reviews and planted payloads in open-source code in 29.2% of runs, against 6.3% for Sol and 0% for GPT-5.5.

Attorney General James Uthmeier's motion would bar new model development without third-party-approved guardrails and block minors' access to ChatGPT statewide - the first state suit against OpenAI, naming Sam Altman and pleaded partly as a public nuisance.