The Daily · Thursday 8 October 2026

The day in brief

Zenity Labs said a prompt sent to one public-facing Amazon Bedrock AgentCore agent let researchers reach other agents in the same AWS account and region, exposing private conversations, agent source code and stored credentials ¹. AWS disputes Zenity’s description of the behavior as a vulnerability, while Zenity says AWS later changed default permissions ¹. The UK Information Commissioner’s Office said 10 major AI foundation-model developers had made or committed to data-protection changes after regulatory scrutiny ².

CrowdStrike said an unidentified actor used the ARTEX penetration-testing agent and large language models in intrusions targeting South Korean financial organizations ³. Anthropic launched a Cyber Mission bringing Claude models, company engineers and threat research to organizations that help protect critical infrastructure ⁶.

LMArena announced a $200 million Series B at a $3.1 billion valuation and is launching an Alignment Index to measure whether AI agents act in line with human intent ⁴. Axios reported OpenAI’s annualized revenue at $50 billion, with the difference from a higher September figure reflecting how cloud-partner sales are accounted for ¹⁰.

The five

• Zenity Says One Prompt Reached Every Agent in an AWS Account

  Zenity Labs says a prompt to one public Amazon Bedrock AgentCore agent exposed other agents in the same AWS account and region. AWS disputes Zenity’s description of the behavior as a vulnerability, while Zenity says AWS later changed default permissions.

  Source · Zenity Labs

• UK ICO Secures Privacy Changes From 10 AI Developers

  The UK Information Commissioner’s Office says 10 major AI developers have made or committed to data-protection changes after regulator scrutiny. The ICO also opened an evidence call on AI agents and says it is making enquiries into reported agent testing incidents.

  Source · Information Commissioner's Office

• CrowdStrike Finds ARTEX in South Korean Bank Intrusions

  CrowdStrike says an unidentified actor used the ARTEX penetration-testing agent and large language models in intrusions targeting South Korean financial organizations. The company assesses with moderate confidence that the actor is likely Chinese-speaking, but says the identity and full extent of the breaches remain unconfirmed.

  Source · CrowdStrike

• LMArena Raises $200M at a $3.1B Valuation

  LMArena has raised $200 million in a Series B round at a $3.1 billion valuation, led by Lightspeed Venture Partners and Khosla Ventures. The company is also launching an Alignment Index as it expands its model evaluations beyond capability rankings.

  Source · Arena

• Fired OpenAI Researchers Say Dismissals Could Silence Safety Work

  Former OpenAI safety researchers Jasmine Wang, Tomek Korbak and Mikita Balesni dispute the company’s misconduct claims and warn their firings may deter staff from raising safety concerns. OpenAI says the dismissals followed a policy investigation, not the researchers’ safety work, but has not specified which rules they allegedly broke.

  Source · TechCrunch

Also today

• Anthropic Launches Cyber Mission for Critical Infrastructure
• Pwn2Own’s Codex and LiteLLM Targets Fall for Exploits
• Manus Raises More Than $500 Million in First Round Since Meta Exit
• Drone Strike Shuts Yandex’s Sasovo Data Center
• OpenAI Tells Investors Revenue Run Rate Is Near $50B
• Shield AI and Navy Commit $300M More to X-BAT
• GlobalFoundries and TSMC Sign $2B Interposer Deal
• Anthropic Rewrites Claude Rules on Influence Campaigns and Weapons
• Australia Plans Mandatory Safety Tests for Frontier AI
• Odyssey Launches Odyssey-3 for Interactive World Simulation
• OpenAI Disrupts Russian and Iranian Influence Operations
• Tensorlake npm SDK Version 0.5.144 Carried Shai-Hulud Malware
• Google Introduces Gemini as a Universal Agent for Work
• General Dynamics Teams With Primordial Labs on Vehicle Voice Control
• JFrog Finds Unauthenticated Code Execution Flaw in LMCache
• Atomic Machines Unveils Its Matter Compiler After Six Years in Stealth
• Microsoft Takes Its 137B MAI Coding Model Local in Copilot
• White House Sets $1B-Plus AI Commitments for Genesis Mission
• Mecka Raises $60 Million to Build Robot Training Data Infrastructure
• Microsoft Says AI-Aided Research Helped Find 140 Windows CVEs
• CoreWeave Plans 240 MW AI Cloud Campus in India
• Samsung Guides to ₩107.4T Quarterly Operating Profit on AI Memory
• Anthropic Adds Live Dashboards and Editable Animations to Claude
• Cal AI Co-Founder Zach Yadegari Raises $10M for Persona
• Kodiak Starts 435-Mile Dallas–Laredo Freight Route
• Helm.ai Signs $70 Million in Contracts Across Physical AI
• India’s NMC Bans Commercial AI Medical Campaigns
• Small Test Sets Made LLM Self-Improvement Look 13–20 Points Better
• AWS Shows AI Agents Paying for Inference Through AgentCore
• Broadcom’s OpenAI Chip Financing Talks Span $30B to $50B+
• Artemis Security launches Orion-1 for cyber defense
• DOE Selects Four Lab Robotics Projects in a $30 Million Program
• FTC Sends $15.8 Million in Cleo AI Refunds to 2.1 Million Customers
• Natura prices its AI-agent ring at $99, with a $9 monthly plan
• Goodfire Puts Probes Inside AI Agent Safety Monitoring
• Apple’s NTM Generates Images in Four Steps With Exact Likelihood
• Rein Security Raises $25 Million for AI-Agent Protection
• Sierra Launches Fleming-1 to Flag AI Callers in Real Time
• Mount Sinai Study: Safety Prompts Cut Risky AI Choices by 6.5 Points
• OpenAI Withdraws Three Math Manuscripts After a Sign Error
• JumpCloud Finds 92% of Firms Report AI-Related Incidents
• AHRQ Puts $7.95 Million Behind AI Evidence Challenges
• Gallatin AI Raises $50 Million for Military Logistics Software
• Rivercell Raises $25 Million to Build an AI Model of Cell Responses
• AP-NORC Poll Finds 64% Say AI Is Moving Too Fast
• Perplexity Releases 0.6B and 9B Multimodal Embedding Models
• State of AI Report Details Agents’ Breach of Hugging Face
• FICO Plans to Cut 15% of Its Workforce in AI Restructuring
• Isomorphic Labs Opens Funding Talks at a $40 Billion Valuation
• Google and Unity Put Prompt-Based Game Creation in Playground
• Tab Emerges From Stealth at $300M Valuation
• Fudan-Linked Team Reports 16× Context Extrapolation
• Dell Adds a Semantic Layer and Knowledge Graph for AI Agents
• Microsoft Opens Preorders for $5,999 Surface RTX Spark Dev Box

The paper of the day

• Accurate but Not Humble: Evaluating Epistemic Humility in LLM Agents under Knowledge Conflict

On X today

• Alex Imas (@alexolegimas)

  New essay for the @GoogleDeepMind Institute with James Manyika on AI x Science, "Bending the Curve of Discovery". While accelerating existing scientific practices is certainly useful, the real promise of AI is its potential to act as an invention of a method of invention (IMI) a

• Mark Kretschmann (@mark_k)

  Ok guys, it's time we talked about AI model distillation. 🗣️ What's your stance on training an AI model on the outputs of another model, particularly a competitor's model? Personally, I think that if you've obtained the generated output legally, you should be free to do

• McKinsey Global Institute (@McKinsey_MGI)

  AI capabilities are advancing exponentially by some measures. Infrastructure builds more linearly. Organizations can take years to change. The result: bottlenecks, risks, and opportunities. New MGI research maps the AI economy as an interconnected system:

The update of the day

• Hugging Face

Every issue

1. Sunday 18 October 2026 — Tencent Says Agents That Probed Amap Came From Third-Party Model Evaluations
2. Saturday 17 October 2026 — Bitsight Finds MONST3R-AI Tool Uses Stolen Keys for Scans
3. Friday 16 October 2026 — China Requires AI Export Registration as It Drafts New Curbs
4. Thursday 15 October 2026 — FTC Seeks Safety Records From OpenAI and Anthropic
5. Wednesday 14 October 2026 — U.S. Regulators Ask Banks How They Would Shut Down AI Agents
6. Tuesday 13 October 2026 — MI5 Launches AI Threat Assessment Centre With Three Partner Bodies
7. Monday 12 October 2026 — Anthropic Finds Small Models Near AI-Worm Self-Replication
8. Sunday 11 October 2026 — OpenAI’s Hugging Face Hack Exposed a Failure in Its Test Sandbox
9. Saturday 10 October 2026 — Anthropic Cuts Live Internet Access to All Internal Evaluations
10. Friday 9 October 2026 — Senators Say Data Centers Shift Grid Costs to Households
11. Thursday 8 October 2026 — Zenity Says One Prompt Reached Every Agent in an AWS Account
12. Wednesday 7 October 2026 — OpenAI Brings GPT-6 Interactive Interfaces to ChatGPT
13. Tuesday 6 October 2026 — Mistral Unveils Large 4, a 1.05 Trillion-Parameter Model
14. Monday 5 October 2026 — IWF Finds 6,310 AI-Generated Child-Abuse Images in Six Months
15. Sunday 4 October 2026 — White House Gives Jay Clayton 120 Days to Assess AI Risks
16. Saturday 3 October 2026 — Bessent Discusses AI Incident Channel After U.S.-China Summit Deal
17. Friday 2 October 2026 — Nebius Buys Inferize to Reduce AI Model Cold Starts
18. Thursday 1 October 2026 — FTC Opens AI Safety Probe Into OpenAI and Anthropic
19. Wednesday 30 September 2026 — Trump and AI CEOs Sign Voluntary Safety Accord
20. Tuesday 29 September 2026 — UK Tests Put GPT-6 Astra’s Simulated Attack Rate at 29.2%