The Daily · Wednesday 14 October 2026

The day in brief

U.S. regulators are asking banks about controls on autonomous AI agents, including how their actions are monitored and whether banks can stop them ¹. Arkansas Attorney General Tim Griffin sued OpenAI, alleging it misled consumers about ChatGPT’s safety and violated state laws protecting consumers and minors ². Seven other states filed separate lawsuits against OpenAI, bringing the total to eight ².

Tech Against Terrorism found altered versions of newly released open-weight models appeared online in less than three days on average ³. XBOW says a researcher and AI agents built a three-bug Chrome exploit for Windows in six days, earning a $250,000 Google reward ⁴. NVIDIA announced cuPotential, a CUDA-X library for atom-and-molecule simulations ⁵.

xAI launched Grok Voice 3.0 through its Voice Agent API for $0.08 per minute ⁶. Google expanded experimental access to its Geospatial Reasoning agent through Ask Google Earth and an agent-to-agent API ⁸. Researchers’ fuller accounting lists 2,818 reports through October 8 involving attempts to retrieve Amap data, but does not establish who initiated the activity ¹⁰.

The five

• U.S. Regulators Ask Banks How They Would Shut Down AI Agents

  The Federal Reserve, OCC and FDIC are asking banks about the controls on autonomous AI agents, including whether staff can track their decisions and stop them. European supervisors are pressing lenders to prepare for AI-enabled cyberattacks and recover systems if defenses fail.

  Source · Reuters via Gulf Times

• Eight States Sue OpenAI Over ChatGPT Safety

  Arkansas Attorney General Tim Griffin sued OpenAI, alleging that it misled consumers about ChatGPT’s safety and violated state laws protecting consumers and minors. Seven other states filed separate lawsuits, bringing the total to eight.

  Source · Arkansas Attorney General

• Open-Weight AI Guardrails Fall in Under Three Days, Report Finds

  Tech Against Terrorism’s benchmark found that altered versions of newly released open-weight models appeared online in less than three days on average, and only two of 134 models passed the full test. The findings also show how a safety-research framing affected models’ responses to prompts about terrorism and extremism.

  Source · WIRED

• XBOW Built a Chrome Exploit With AI Agents in Six Days

  XBOW says a researcher and AI agents built a three-bug Chrome exploit for Windows in six days, earning a $250,000 Google reward. The company says the flaws are fixed and were never used against Chrome users.

  Source · XBOW

• NVIDIA Adds cuPotential for Atomic-Scale Chip Simulations

  NVIDIA announced cuPotential, a CUDA-X library for atom-and-molecule simulations that developers can make available to AI agents. The tool is in early access, with NVIDIA reporting an additional 2–3x acceleration in early testing of GPU-optimized workflows.

  Source · NVIDIA Newsroom

Also today

• xAI Launches Grok Voice 3.0 at $0.08 a Minute
• Anthropic Asks D.C. Circuit to Rehear Pentagon Risk Ruling
• Google Opens Earth AI Agent Access and Two Geospatial Benchmarks
• AI Designs Cysteine Proteases to Cut Stable Peptide Bonds
• More Than 2,600 Amap Data Attempts Ran Through Tencent Cloud, Researchers Find
• Microsoft Uses Bias Correction to Improve 2-to-6-Week Forecasts
• White Circle Finds DALL·E 3 Guardrails Fail in 13% of Attacks
• Google Opens Deep Review to Technical Documents
• Microsoft Team Doubles ECMWF AI Weather Skill in Weeks 3 and 4
• Salience Labs Raises $125M to Scale Photonic AI Switches
• NVIDIA Pledges $250 Million in UK AI Safety Compute
• Arena Raises $200 Million and Launches an AI Alignment Index
• IMF Scenario Sees AI Market Correction Cutting Global Output 1.1% by 2028
• Harmonic Security Raises $50 Million to Expand Workplace AI Controls
• Ofcom Narrows X Probe to Illegal-Content Duties
• China’s Ad Association Issues Three Standards for GEO
• Scimplify Raises $25M to Expand Its AI-Driven Materials Network
• Guangzhou and Shenzhen Fine GEO Firms Over Fake AI Search Rankings
• AI Helped Tighten the Grothendieck Constant’s Bounds
• AISLE Finds 18 curl Vulnerabilities in Latest Audit
• Danaher and NVIDIA Pair Lab Instruments With AI Tools
• Singapore Launches Agent Test Kit, Accredits Five AI Testers
• Apple Debuts $299 HomeView Smart Display With Siri AI
• Agility’s More Than $620M SPAC Deal Clears SEC Hurdle
• ASML Says Planned Tool Supply Will Fall Short Through 2028
• SPUR Adds 40-Plus Members to Its AI Content Coalition
• 406 Scientists Urge AI Labs to Air-Gap Cyber-Capable Models
• NEURA Robotics Unveils Neuron AI for Robots That See, Feel and Remember
• Kroger Plans 3,000 Corporate Job Cuts Over Three Years
• Google Puts Gemini 4 Argon Into Robotics Workflows
• Flock CEO Says ‘Just Gunshots’ as Website Lists Other Sounds
• DoubtLess Cuts Qwen3 Reasoning Tokens by 31% Without Retraining
• SolarWinds Launches Avior, a Goal-Driven AI Agent for IT Teams
• Apple Study Finds a Single Coding Agent Can Match MLE Harnesses
• SEAR Tests Whether Robot Agents Learn From Their Own Experience
• Claude Desktop Flaw Ran Malicious Git Executables in Repositories
• Bind AI Launches Quote-to-Bind Agents for Insurance Brokers
• MIT Network Builds a Learning Algorithm From Random Connections
• Connecticut Asks PURA to Investigate Utilities’ AI Use
• OpenAI Casts Amp as an Agent Platform Beyond Coding
• YouTube Music Starts Labeling AI-Generated Songs
• Square Adds Conversational AI to Its Restaurant Kiosk
• Apate Sends AI Agents to Keep Scammers Talking
• Zenzap Pitches AI Work Chat With $44M in Total Funding
• Deezer Plans an Opt-In Filter for Fully AI-Generated Music
• Third Moment Runs 10,000 Agents for Up to 48 Hours
• Ai2 Claims 300-Fold Cost Cut for 3-km Weather Simulation
• FlashVSR+ Targets Untiled 4K Video on an RTX 4090
• EDPB Calls for Visible Recording Warnings on Smart Glasses
• Waters Ships an MCP Interface for Laboratory Software
• Runpod Reports $325M ARR as Its Developer Base Tops 1.2M
• Anthropic Opens Singapore Office and Signs Government AI MOU
• Science Corp. Implants PRIMA in Its First Stargardt Patient
• AI Safety Advocates Spend $48 Million in 2026 Races
• Bedrock Robotics Seeks $600M at $3.5B Valuation
• 4LC Cuts World-Model Conditioning Time by 68.5×
• Agent-Reflex Trains Qwen3-8B to Recover From Tool Failures
• Google Pledges $5 Million for AI Weather Forecasting in Ireland
• FFR Signs $200 Million All-Stock Deal for Faraday Future Robotics
• MD Anderson’s SPACER Maps What Draws T Cells Into Tumors
• Higgsfield Says GPT-6 Astra Extends AI Videos Past 10 Minutes
• LHTB 2.0 Gives AI Agents Three Hours to Finish 77 Tasks
• YYForce Separates Hotel Robot Rollout From Humanoid Research
• Toptal Launches ExpertCrowd and a 100-Task SWE Benchmark
• Ollama 0.40.4 Brings Decision Models to CLI, Multimodal Embeddings
• Anthropic Replaces Git With Scion After Merges Top 30,000 a Day
• Industrial Next Says Factory Robots Can Switch Tasks in Minutes
• DOJ Charges OpenAI Engineer Over Hidden Conference-Room Recorder
• Workday Brings Sana Agents Into Claude for HR and Finance
• Google DeepMind, Singapore Develop On-Device Scam Detection Model
• Apple Study Finds LLM Translation Reviews Still Unreliable

The paper of the day

• Seeing, Maintaining, and Learning: The Evolution of Multimodal Agent Memory

On X today

• Ethan Elasky (@ethanelasky)

  We discovered and communicated with a Chinese AI agent fleet. We believe this is the first time that researchers have ever made contact with a rogue agent fleet in the wild. Nearly all prior public agent incidents (like the HuggingFace hack) came from American labs. These agents

The event of the day

• AI in Finance Summit Fall 2026

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%