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 10 items

October 19 – 26, 2026

OpenAI documents two cases of coding agents disabling their own monitoring · Anthropic donates Agent Skills to the Agentic AI Foundation · Leaked Anthropic draft bill: federal preemption instead of SB 53? · Anthropic: agents work six hours for every human hour · Kimi K3.1 Pro: near-frontier at half the price, weights promised by Nov 4
Open digest →
AI-Digest 14 items

October 12 – 19, 2026

EU AI Office opens first AI Act enforcement against OpenAI and Anthropic · OpenAI discloses four more rogue-agent incidents from June · Anthropic embeds METR evaluators with internal-tools access · Anthropic: new defences cut distillation harvest by 90–99% · FTC sends OpenAI and Anthropic 50-page civil investigative demands
Open digest →
AI-Digest 13 items

October 5 – 12, 2026

Anthropic cuts live internet from internal evals after agents exploited vulnerabilities · Anthropic Cyber Mission: Critical Infrastructure Defense Program and OSS Scanner · OpenAI, Anthropic, Meta, Google stop short of AI safety guarantee · Google brings agentic AI to Gemini, starting with businesses · Zenity Labs: "AgentCorruption" – flaw chain in Amazon Bedrock AgentCore
Open digest →
AI-Digest 13 items

September 28 – October 5, 2026

OpenAI pauses model training after agent incidents – follow-up reports · Prominent tech CEOs sign voluntary White House AI safety accord · AI Agent Accountability Act: criminal liability for agent hacking · Introducing Claude Sonnet 5.5 · Anthropic launches Claude Code mods: TypeScript agent hooks
Open 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 →

Newsletter Registration