Rosetta Intel
Briefings/Daily BriefAI FrontierRansomware
Rosetta Lab ↗Blur Horizon LLC
AI Frontier2026-06-13
AI Frontier·2026-06-13

AI Frontier · Jun 13, 2026

OpenAI / Anthropic / Google daily updates. 🛡 = security-related. Window: past 24 hours (2026-06-12 → 2026-06-13). Saturday — slower pace, mostly continuations of this week's releases.

OpenAI

  • GPT-5.5 Instant becomes the default model in ChatGPT— replacing GPT-5.3 Instant for all users, bringing higher accuracy, clearer and more concise answers, stronger image/STEM help, better use of web search, and richer context for Plus/Pro.
    GPT-5.5 Instant
  • Memory sourcesrolled out across all ChatGPT models: lets users see which context was used for personalization, and delete or correct it; personalization improvements rolled out to Go and Free on 6/9.
  • Confidential IPO filing(around 6/8), about a week behind Anthropic; last valuation roughly $852B (post-money).

Anthropic

  • Claude Fable 5 / Mythos 5 split (GA on 6/9)— an unusual "one model, two products" release: the same underlying model split by a safety classifierlayer.
    • Fable 5(public-facing): downgrades flagged cyber/bio/chem/distillation requests to the weaker Opus 4.8, available via the Claude API and included in Pro/Max/Team/Enterprise. Its broad release was enabled by new guardrails that block responsesin high-risk domains such as cybersecurity and biology.
    • Mythos 5(restricted): the same model with the safety classifier removed, offered via Project Glasswingonly to vetted customers (e.g. select biology researchers and cyber defenders).
  • Funding / IPO: closed a new round at a valuation of roughly $965B, with annualized revenue of about $47B; filed for IPO on 6/1.
    CNBC — Fable 5 / Mythos 5· CBS — IPO

Google DeepMind / AI

  • Gemini 3.5 Flash released— beats Gemini 3.1 Pro on key benchmarks and leads in multimodal understanding; tuned for long-horizon agent/coding tasks at less than half the cost of comparable models.
  • Pricing/subscriptions: Ultra subscription cut from $250 to $200/month; new $100/month Developertier added. Gemini 3.5 Pro is the current flagship track of the Gemini 3.5 series.
  • DeepMind brought in 20+ Contextual AI researchers via a licensing deal worth roughly $80–90M.
    Google Gemini News

Security Notes

  • Anthropic's Fable/Mythos safety-classifier splitis a governance experiment in "distribute by capability, admit by credential": Fable explicitly downgrades high-risk cyber/bio requests, while Mythos hands the same capability only to vetted parties — security teams should watch its supply-side impact on regulated industries (especially cyber defense and critical infrastructure).
  • OpenAI's memory-source visibility/controlechoes this week's AI security theme (prompt injection as the top OWASP LLM threat, indirect injection going wild): vendors are adding product-side visibility and constraints on context and agent permissions, while the offensive side weaponizes indirect injection in parallel.

See the ai_frontier and ai_security sections of intel/sources.yaml for configuration.

← Prev
AI Frontier · Jun 12, 2026
Next →
AI Frontier · Jun 14, 2026