Morning Edition · Friday, October 9, 2026
Tech
OpenAI Withdraws Three of 722 AI-Written Math Papers as Mathematicians' Group Urges Boycott
A sign error invalidated one proof and the two papers that depended on it, and Lean formalizations so far cover the main result of only 162 of the 722 manuscripts.
Tech
Anthropic Prices Claude Haiku 5.5 at $0.10 per Million Input Tokens, Matching OpenAI's GPT-6 Luna
The cut is 90% from Haiku 4.5's list price, and it leaves Anthropic and OpenAI at identical rates in both the small-model and mid-tier brackets.
Tech
Nvidia and Microsoft Open Preorders for RTX Spark Laptops With Up to 128 GB of Unified Memory for Local Agents
Microsoft also made Execution Containers generally available, so Windows itself now enforces what files and network access a background agent can use.
Tech
Two Security Papers Find Agent Defenses Miss Attacks: Skill Scanners Bypassed 94% to 100% of the Time, Triage Agents Miss 40% of Alerts
One attack hides malicious Python bytecode behind benign source code, and the other study finds that agents dismiss attack alerts when their own searches return no records.
Geopolitics
OpenAI Bans Russian and Iranian Influence Networks That Used Fake Journalists and a Latin American Think Tank
OpenAI rated the Russian operation Category 5 on a six-point scale, the first time it has disrupted an operation at that level.
Tech
Anthropic's New Usage Policy Bars Claude From Recommending Whom Police Should Investigate or Arrest
The rules take effect November 12, and they also ban tracking people without consent, add safety requirements for autonomous hardware and prohibit sustained abuse of the model.
Tech
New Statistical Certificate Shows Separated Leaderboard Rankings Withstand About 2,000 Forged Votes
The method publishes an attack budget with each pairwise ranking claim, and the guarantee holds at all times even against attackers who watch every published standing.
Tech
Agent-Controlled Forgetting Cuts a Debugging Session's Prompt Tokens From 912,000 to 232,000 Without Retraining
A separate paper finds that a widely cited gain from encoder-only healing in SVD KV-cache compression disappears once each method gets its own tuned learning rate.
Tech
AI Agents Now Make Up About Half of Visits to the St. Louis Fed's FRED Database
Federal Reserve Governor Christopher Waller said visits are growing about 150% a year, mostly from bots, and the St. Louis Fed has opened a Model Context Protocol connector to its data.
Tech
Meta's SAM 3 and DINOv3 Cut Beamline Image Analysis at Five US National Labs From a Month to About 15 Minutes
Meta says the open vision models run on 300 A100 GPUs inside the Energy Department's Genesis Mission, and the department announced new Genesis awards on Thursday.
Tech
Judge Sentences Michael Smith to 18 Months in First Federal AI Music Streaming Fraud Case
Prosecutors had asked for at least 46 months for the scheme, which used hundreds of thousands of AI-generated songs and up to 10,000 bot accounts to collect more than $8 million in royalties.
Tech
Anthropic Offers Startups a Free Year of Claude Team and $1,000 in API Credits
Founders backed by venture firms in Anthropic's partner network can get up to $100,000 in additional credits, and Meta is pursuing the same developers with a $20 sign-up credit and a lower price for customers who share their data.