Morning brief

Sunday, June 14

1 people · 0 companies · 5 topics

TL;DR The US government forced Anthropic to pull Fable 5 and Mythos 5 three days after launch, citing a jailbreak; June 15 tomorrow brings Claude 4 retirements and an Agent SDK billing split that will break existing workflows.

Models & research

  • Anthropic received a government directive June 12 (5:21pm ET) blocking foreign nationals from Fable 5 and Mythos 5; Anthropic pulled both models for all users globally to comply, publicly disputes the call, and says the jailbreak found is minor and present in other public models too. Anthropic statement · Bloomberg · Time · Decrypt
  • Fable 5 had launched June 9 as a Mythos-class model made safe for general use; it was free for Pro/Max/Team/Enterprise users until June 22, now suspended globally. TechCrunch
  • Moonshot AI dropped Kimi K2.7-Code weights on Hugging Face June 12: 1T-param MoE (32B active, 384 experts), 256K context, Modified MIT license, +21.8% on Kimi Code Bench v2 vs K2.6 with 30% fewer reasoning tokens. MarkTechPost · HuggingFace

People

  • simonw wrote up the Anthropic Fable ban on June 13, linking the official statement; one of the first practitioner commentaries noting how unprecedented a government commercial model recall is. simonwillison.net

Tools & dev

  • claude-code two deadlines hit June 15 tomorrow: claude-sonnet-4-20250514 and claude-opus-4-20250514 retire with no grace period (API calls will error immediately), and Agent SDK usage splits into its own billing pool billed at standard API rates separate from subscriptions. Claim your credit before midnight. Enterprise DNA · UsageBox
  • Microsoft is canceling claude-code licenses across its Experiences+Devices division by June 30, steering roughly 5,000 engineers to GitHub Copilot CLI after token costs blew the annual AI budget; internal devs are openly unhappy about the feature gap. OpenTools · Windows Central

OSS

  • NVIDIA SkillSpector is today’s top trending GitHub repo (962 stars today): a security scanner for agentic-ai skills that detects 64 vulnerability patterns across 16 categories, supports OpenAI, Anthropic, and NVIDIA models. GitHub
  • OpenHands (AI-driven dev agent) is trending with 258 new stars today. GitHub
  • LMCache is trending (271 new stars): a KV cache layer for LLM inference reporting up to 15x throughput improvement for long-context and RAG workloads when combined with vLLM. GitHub