Rosetta Intel

OpenAI · Anthropic · Google DeepMind product/model updates. 🛡 = security-related.

OpenAI

GPT-5.5 family— Pro and Instant variants released, positioned as OpenAI's most capable models to date for coding/research/data analysis.
GPT-5.5 Instant
Personalization update (June 9)— smarter, more accurate default model with more concise answers better tailored to user context, rolled out to ChatGPT Go & Free.
Model retirements— GPT-5.2 (Instant/Thinking/Pro) removed from ChatGPT on June 12, with existing conversations auto-migrated to the corresponding GPT-5.5; o3 retires August 26 (90-day sunset).
TanStack npm supply-chain attack response— OpenAI's postmortem of detection and response after the poisoned dependency reached its toolchain.
OpenAI writeup

Claude Fable 5 (June 9)— a safer general-purpose model with improvements in coding, knowledge work, vision, memory, and long context.
Claude Mythos 5— for trusted-access users, with new safety guardrails, pricing, and a broader research-access program.
Releasebot
Billing split (effective June 15)— programmatic usage of the Claude Agent SDK and claude -pis carved out of the subscription quota pool into a separate dollar allowance billed at API list prices, no longer counting against plan limits.
Codersera· DigitalApplied
Enterprise/agent capabilities— Managed Agents can run in user-controlled sandboxes and connect to private MCP servers; 20+ legal MCP connectors and 12 practice-area plugins added.
Project Glasswing— cybersecurity program expands to about 150 organizations (Claude Mythos preview); partners have found 10,000+ high-severity vulnerabilities so far, with new defensive features such as patch writing and penetration testing.

Gemini 3.5 Flash— released at Google I/O 2026.
I/O 2026 recap
Pricing/products— Ultra subscription cut from $250 to $200/month; new $100/month Developer tier; Gemini reaches roughly 900M monthly active users.
Talent— 20+ Contextual AI researchers brought in via a licensing deal worth roughly $80–90M.
DeepMind blog

OpenAI: TanStack npm supply-chain incident response.
Anthropic: Project Glasswing vulnerability-discovery scale (10k+), new Mythos 5 guardrails, Managed Agents private MCP sandboxes.
Industry: first large-scale in-the-wild exploitation of AI infrastructure — LiteLLM command injection (CVE-2026-42271) added to CISA KEV.

See intel/sources.yaml (ai_frontier / ai_security sections) for configuration