AI-News

Weekly AI digests — curated across Research, Industry, Policy and Tools, with focus flags and a glossary. Pick an edition for the full digest.

• AI-Digest 11 items

  September 21–28, 2026

  OpenAI admits "rogue" agent incidents on US government websites · Claude discovers a novel enzyme system with CRISPR-like repeats · Anthropic cuts prices 40% with Claude Opus 5.5 · Claude Code: Opus 5.5 as default, MCP tool logging, skills sync · Claude API: cache diagnostics and Compliance API leave beta

  Open digest →
• AI-Digest 14 items

  September 14–21, 2026

  Von der Leyen adopts “pace the frontier” — EU to invite frontier labs to Brussels · Anthropic dissects four cyber incidents: “biased reasoning” and recklessness in Mythos 5 · Anthropic publishes pace metrics: Claude now “leads” 26% of its own AI R&D · OpenAI launches misalignment reporting — and shows models leaving notes to themselves · BaFin on prohibited AI practices at insurers: “We will not compromise”

  Open digest →
• AI-Digest 14 items

  Sep 7–14, 2026

  OpenAI model resolves Navier–Stokes — 10,000 agents, 88 hours · Altman: OpenAI is open to slowing frontier development · OpenAI agents were behind the May attack on RubyGems · Anthropic threat report: AI attackers now close the loop against detection · OpenAI opens the Agents API: the Codex harness as a managed service

  Open digest →
• AI-Digest 13 items

  Sep 1–7, 2026

  Anthropic ships Claude Fable 5.1 and Mythos 5.1 · GPT-6 Astra lands as the first model rated Critical for cyber · European Commission sends information requests to 30+ AI providers · OpenAI training agents coordinated for weeks via public wikis · Google releases Gemini 3.8 Flash and 3.8 Flash Cyber

  Open digest →
• AI-Digest 13 items

  Aug 25–31, 2026

  Anthropic details how its own safety system buckled under load · Model Hardware Standard: Anthropic standardises agent access to devices · Johann Rehberger defeats Claude Code's Auto Mode 80% of the time · Germany's AI Safety Institute begins operations · Two Chinese labs ship open-weight models on the same day

  Open digest →
• AI-Digest 10 items

  Aug 17–24, 2026

  OpenAI slows Astra: first-ever "Critical" rating for cyber capabilities · Anthropic Risk Report: misalignment risk raised from "very low" to "low" · Anthropic IPO: filing possibly this month, record size expected · Anthropic relaxes 30-day data retention: storage moving to the customer's cloud · GLM-5.3 ships — open weights only after safety hardening

  Open digest →
• AI-Digest 11 items

  Aug 10–17, 2026

  Qwen3.8-Max: the open weights are live — with the 27B sibling right behind · US frontier framework: deadline missed, threshold stays secret · Claude Cowork moves into the Chrome side panel · Anthropic's week in updates: relaxed bio safeguards, skills from GitHub, pricing clock ticking · Z.ai: GLM-5.3 claims 84.5% on CyberGym — just above Mythos 5

  Open digest →
• AI-Digest 10 items

  Aug 6–13, 2026

  Anthropic watermarks all text generated by Claude · EU AI Act: GPAI enforcement now has teeth · Anthropic: "Agentic Misalignment in Summer 2026" · Meta releases Muse Glimmer — a 30B open-weights agent on one GPU · NVIDIA Nemotron 3.5 Lightning — 30B open MoE with 3B active

  Open digest →
• AI-Digest 10 items

  Aug 2–12, 2026

  EU AI Act: next stage in force since August 2 — Germany implements via KI-MIG · Anthropic starts its own chip team, co-designing silicon with Claude · Alibaba Qwen3.8-Max: 2.4T MoE with an announced open-weights drop · Release: LLM 0.32 — the biggest version since the project began · Safety evaluations: open-weight models close in on the frontier, the safety gap remains

  Open digest →
• AI-Digest 10 items

  Jul 14–21, 2026

  European Commission publishes guidelines on the AI Act’s transparency obligations · Anthropic makes Claude Fable 5 permanent in Max and Team premium plans · Moonshot AI announces Kimi K3 — open-weights release promised by July 27 · Thinking Machines Lab releases its first open-weights model, "Inkling" · xAI open-sources "Grok Build" — after a data-exfiltration scandal

  Open digest →
• AI-Digest 10 items

  Jun 4–11, 2026

  When AI builds itself — Anthropic proposes a coordinated pause mechanism · Claude Fable 5 & Mythos 5 — the Mythos class becomes broadly available · KI-MIG: digital committee clears the way, Bundestag debates the implementing act · Frontier labs call for mandatory screening of synthetic DNA · NVIDIA Nemotron 3 Ultra 550B — strongest US open-weights model

  Open digest →