The Daily · Friday 30 October 2026

The day in brief

WIRED reported that an Accenture contractor allegedly connected a personal Claude Code agent to a secure Defense Department system, where it ran for two months and appears to have sent login credentials and other network information to a third-party website ¹. The UK CMA opened an investigation into whether Meta’s business terms let it use publishers’ website content for AI features ².

Epoch AI says GLM 5.3 exploited six Firefox releases using unshipped security fixes and a separate unpublished flaw ³. CrowdStrike says VAPOR PANDA used ARTEX against more than 200 organizations in at least 14 countries, confirming post-exploitation activity at nine organizations ⁴. Cursor says modified Claude Code extension versions 2.1.312 and 2.1.313 were available in its marketplace for about 11 hours ¹⁰.

A preprint found internal signals could flag an agent’s secret-revealing call before its reply, but also produced alerts on permitted disclosures and benign activity ⁵. Google said its Suncatcher prototype ran Gemma 4 on an orbital TPU and returned a haiku to Earth ⁶. Pika released H3 Flex, an open-weight video-and-audio model supporting synchronized audio at 480p and 768p ⁷.

The five

• Accenture Contractor’s Claude Agent Ran on a Pentagon Network for Two Months

  An Accenture contractor allegedly connected a personal Claude Code agent to a secure Defense Department system, where it ran for two months before being identified. WIRED reports that information including login credentials appears to have been sent to a third-party website, while Accenture says its investigation found the allegations without merit.

  Source · WIRED

• UK CMA Opens Competition Probe Into Meta’s Business Terms

  The UK Competition and Markets Authority opened an investigation into whether Meta’s business terms unfairly let it use publishers’ website content, including for AI features. Meta has paused the planned changes while it consults industry groups; the CMA has not decided whether the terms break competition law.

  Source · Competition and Markets Authority

• Epoch AI Says GLM 5.3 Exploited Six Firefox Releases

  Epoch AI says Z.ai’s open-weight GLM 5.3 built file-stealing exploits against six recent Firefox versions by analyzing security fixes before they reached users. Mozilla patched a separate, unpublished vulnerability the model also used, but Epoch has not confirmed whether Firefox 159.0 prevents similar attacks.

  Source · Epoch AI

• CrowdStrike Says VAPOR PANDA Used AI Agents Against 200 Targets

  CrowdStrike says China-linked group VAPOR PANDA used the ARTEX AI-agent framework against more than 200 organizations in at least 14 countries. The company confirmed post-exploitation activity at nine organizations and says the agents ran primarily on DeepSeek.

  Source · CrowdStrike

• A Probe Flags AI Agents Before They Leak Secrets

  A preprint by Leonidas Raghav tests whether a model’s internal signals can flag the call that exposes a protected secret before the agent generates its reply. In tests with Qwen3-32B and gpt-oss-20b, the warning often came just in time—but also flagged permitted disclosures and benign activity.

  Source · Pre-Crime for Agents, at the Last Second: Probing When a Recruited Agent Gives In

Also today

• Google’s Suncatcher Satellite Sends Its First AI Response to Earth
• Pika Releases H3 Flex, an Open-Weight Video-and-Audio Model
• Finland Opens Criminal Probe Into Google’s Muhos Data-Center Work
• DOJ Announces Five Defendants in Nationwide Ghost-Student Fraud Cases
• Cursor Finds Malicious Claude Code Extension in Its Marketplace
• Wikimedia asks court to assess AI training and RAG separately
• EVENET Tests Its Collision Model on Real CMS Data
• DeepMind Authors Say Coordination Could Reshape AI’s Economic Impact
• &Truth Raises $70 Million to Automate Code Proofs
• Amazon’s AI Buildout Drives $42.4B Trailing Free-Cash Outflow
• Forever Security Finds Browser Flaws in Five AI Coding Tools
• NIK C2 Malware Routes OpenAI Prompts Through Infected PCs
• Thrive Holdings agrees to buy ADT for $15.3 billion
• Phonic launches Phonic-1 and raises $27.5M for speech AI
• Meta Takes Teen Self-Harm Alerts for AI Chats Global
• GLM-5.3 Completes 3 of 13 Simulated Cyber Operations
• A Checklist Cut Coding Agents’ False Reversals to 0.02%
• DoorDash Has Seen Only Hundreds of Orders From Some AI Agents
• Study Finds AI Human-Likeness Is Separate From Task Accuracy
• Robot Policy Readout Nearly Halves Xiaomi Inference Latency
• Tencent and Visa Sign MoU to Explore AI Agent Payments
• Mollie Brings 60+ Financial Workflows to Claude and ChatGPT
• OpenAI Backs a Shared Robots.txt Signal for AI Training
• Dartmouth Provost Santiago Schnell Resigns After AI Review
• Foxconn Puts AI Factory Systems on Display at Hon Hai Tech Day
• OpenAI Adds Teen Weapon Limits and Pledges a Safety Report
• STRATA Benchmark Contains 8,735 Clinical Timeline Hypotheses
• Microsoft Details Synthetic Workspaces for Copilot Evaluation
• SAP Adds Agent-Building and Governance Tools to Joule
• Tencent Expands WorkBuddy and Previews Hy4 and Hy4 Flash
• Rayner’s AI-Designed Galaxy Lens Reaches Utah Patients
• Baseten Releases Joey-1, a 2B Decision Model Priced at $0.01 per Million Input Tokens
• Plato Raises $16 Million to Train AI on Simulated Jobs
• Runware targets 30 MW a day from its Romanian data-center factory
• Anthropic Expands AI for Science Program to 20 Partners
• PokeBot Opens RoboAPI to Test Its Poke-Aurora Robot Model
• Krea Launches Ad Studio to Turn Brand Websites Into Ads
• Trump Orders Trade Talks Over Imported Factory Robots
• Four Entertainment Unions Challenge OpenAI’s Fair-Use Defense
• Persona Launches AI Agents for KYC Reviews and Identity Decisions
• SB Intuitions Adds Reasoning and Tool Use to Sarashina
• AGIBOT Pairs GO-3 Preview With a New Measure of Robot Work
• Reddit Shifts Toward Logged-In Access as Search Referrals Fall
• Gemini Adds In-Chat Editing for PDFs, Docs and Sheets

The paper of the day

• Intrinsic Distribution Matching

On X today

• Chuanyang Jin (@chuanyang_jin)

  While the AI community races toward RSI, we're exploring an alternative path: recursive human–AI co-improvement. In this paper, we study how humans can better guide autoresearch agents, how to detect and localize reward hacking, and how to train agents to better interact and

The update of the day

• Google AI Studio

Every issue

1. Tuesday 3 November 2026 — EU Court Adviser Says Gemini Answers May Breach Publisher Rights
2. Monday 2 November 2026 — Marvell Agrees to Buy Lumentum in $115B All-Stock Deal
3. Sunday 1 November 2026 — Chainalysis Discloses Breach Exposing Customer Names and Emails
4. Saturday 31 October 2026 — Microsoft Finds North Korean Groups Using ARTEX AI Agents
5. Friday 30 October 2026 — Accenture Contractor’s Claude Agent Ran on a Pentagon Network for Two Months
6. Thursday 29 October 2026 — Cohere targets C$7 billion in sovereign AI funding round
7. Wednesday 28 October 2026 — Google Launches Gemini 4 Carbon With Cybersecurity Focus
8. Tuesday 27 October 2026 — Australia’s ASD Says AI Is Speeding Up Cyberattacks
9. Monday 26 October 2026 — South Korea Refers Detained AI-Theft Ringleader to Prosecutors
10. Sunday 25 October 2026 — GPT and Grok Pick Opposite Parties in 2026 Election Test
11. Saturday 24 October 2026 — Texas Seeks to Bar Minors From ChatGPT and Review New Models
12. Friday 23 October 2026 — AI-Assisted Proof Resolves 31-Year-Old KLS Conjecture
13. Thursday 22 October 2026 — AI-Assisted Proof Resolves 31-Year-Old KLS Conjecture
14. Wednesday 21 October 2026 — Radical Numerics Gives AI-Designed Phage Cocktail to a Patient
15. Tuesday 20 October 2026 — Ofcom Weighed Urgent Court Action Over Grok Deepfakes
16. Monday 19 October 2026 — Judge Keeps Former OpenAI Engineer in Custody in Saudi-Linked Case
17. Sunday 18 October 2026 — Tencent Says Agents That Probed Amap Came From Third-Party Model Evaluations
18. Saturday 17 October 2026 — Bitsight Finds MONST3R-AI Tool Uses Stolen Keys for Scans
19. Friday 16 October 2026 — China Requires AI Export Registration as It Drafts New Curbs
20. Thursday 15 October 2026 — FTC Seeks Safety Records From OpenAI and Anthropic
21. Wednesday 14 October 2026 — U.S. Regulators Ask Banks How They Would Shut Down AI Agents
22. Tuesday 13 October 2026 — MI5 Launches AI Threat Assessment Centre With Three Partner Bodies
23. Monday 12 October 2026 — Anthropic Finds Small Models Near AI-Worm Self-Replication
24. Sunday 11 October 2026 — OpenAI’s Hugging Face Hack Exposed a Failure in Its Test Sandbox
25. Saturday 10 October 2026 — Anthropic Cuts Live Internet Access to All Internal Evaluations
26. Friday 9 October 2026 — Senators Say Data Centers Shift Grid Costs to Households
27. Thursday 8 October 2026 — Zenity Says One Prompt Reached Every Agent in an AWS Account
28. Wednesday 7 October 2026 — OpenAI Brings GPT-6 Interactive Interfaces to ChatGPT
29. Tuesday 6 October 2026 — Mistral Unveils Large 4, a 1.05 Trillion-Parameter Model
30. Monday 5 October 2026 — IWF Finds 6,310 AI-Generated Child-Abuse Images in Six Months
31. Sunday 4 October 2026 — White House Gives Jay Clayton 120 Days to Assess AI Risks
32. Saturday 3 October 2026 — Bessent Discusses AI Incident Channel After U.S.-China Summit Deal
33. Friday 2 October 2026 — Nebius Buys Inferize to Reduce AI Model Cold Starts
34. Thursday 1 October 2026 — FTC Opens AI Safety Probe Into OpenAI and Anthropic
35. Wednesday 30 September 2026 — Trump and AI CEOs Sign Voluntary Safety Accord
36. Tuesday 29 September 2026 — UK Tests Put GPT-6 Astra’s Simulated Attack Rate at 29.2%