
A bipartisan group of states have also individually sued OpenAI
Iowa Attorney General Brenna Bird announced at a press conference in Des Moines that she is suing OpenAI for hurting Iowa children, deceiving parents, and violating Iowa’s Consumer Fraud Act, her office said. A bipartisan group of states have also individually sued OpenAI, the office said. Kansas Attorney General Kris Kobach, Nebraska Attorney General Mike Hilgers, and West Virginia Attorney General JB McCuskey joined Bird at the press conference, and Arkansas, Arizona, Idaho and Montana have also sued OpenAI in their own state courts, according to the office.
Read the full story
Business

Business
YouTube Music announces labels flagging songs made with generative AI
Compute & Infrastructure
NVIDIA adds cuPotential library for atomic simulation in early access
In early-access testing, the library accelerated already GPU-optimized workflows by an additional 2-3x, the company said.
Models & Tools
OpenAI alleges Apple hire sent more than 100 unreleased product files
OpenAI says in a new court filing that its secrets are now circulating inside Apple, and it may file a countersuit.
Ai2 introduced HiRO-ACEv2, the newest update to its High Resolution Output Ai2 Climate Emulator (HiRO-ACE)
Ai2 says it cannot yet produce high-resolution data for other climates.
Research & Evaluation
Authors may list the tools a skill pre-approves in its own frontmatter
Seven listings in ten carry at least one elevated capability, and almost none of it is visible at install time.
Greater argmax agreement between student and teacher does not consistently improve accuracy across model pairs
Apple reports error rate reductions in code-switched speech recognition
The research reports error rate reductions of 6.35 percent on the SEAME devman subset and 8.29 percent on devsge.
Business
OpenAI infrastructure lead Sachin Katti departs after roughly a year
Uday Ruddarraju, OpenAI's CTO of compute, will replace him and lead the compute organization.
Taiwan Semiconductor Manufacturing Co.
Its operating margin reached 58.7%, above the upper end of its guidance range of 58%.
Policy & Law
OpenAI did not immediately move to eject Alfozan
The complaint alleges a chatbot account tied to the engineer's spouse submitted prompts about building and installing such a device.
Briefing
Michael Sullins and one other author submitted a paper titled IR-PRM+ with Previous-Utility Prediction: Constant Regret and Best-Iterate Convergence on 13 October 2026
Michael Sullins and one other author submitted a paper titled IR-PRM+ with Previous-Utility Prediction: Constant Regret and Best-Iterate Convergence on 13 October 2026. Regret matching and its variants are central in practice for solving normal-form and extensive-form games, but have historically lacked strong theoretical convergence guarantees. The authors report that previous-utility prediction recovers constant-regret and convergence-rate guarantees for both simultaneous and alternating IR-PRM+ updates in finite two-player zero-sum matrix games. They also report the current duality gap vanishes and the current profile converges to the Nash equilibrium set, with analysis extended to extensive-form games via the IR-PCFR+ variant.
Briefing
- 02
The heads of the Bank of England and the IMF warned that rising cyber-attacks risk undermining trust in the financial system
The Guardian reported that the heads of the Bank of England and the IMF warned that rising cyber-attacks risk undermining trust in the financial system. At IMF and World Bank meetings in Bangkok, Andrew Bailey, governor of the Bank of England, pointed to alarming headlines about self-replicating AI agents hijacking personal devices and a French supermarket hack that heated fridges up to 50C. Bailey said that for him the heart of the debate about AI is trust in technology and the human control needed to achieve that trust. He also said: "If we don't have trust in money, we don't operate as a society." The IMF had warned that a big cyber-attack was one factor raising the global financial stability threat level to "elevated".
- 03
The FBI's Los Angeles division distributed a memo in May 2026 warning that threat actors will likely increase their use of generative AI tools for attack planning
The FBI's Los Angeles division distributed a memo in May 2026 warning that threat actors will likely increase their use of generative AI tools for attack planning, the Guardian reported. The Guardian described the warning as following a May 2025 car bombing outside a Palm Springs fertility clinic, in which a 25-year-old man killed himself and injured four others. According to the memo, the suspect used an unnamed generative AI tool that failed to detect his request for bomb-making instructions, and the privacy of the chat hindered alerting police. The memo states: "Current built-in GenAI safeguards are insufficient to prevent illicit use." The Guardian reviewed more than 80 documents, obtained by the transparency non-profit Property of the People, that law enforcement distributed between 2021 and 2026.
- 04
Humanly acquires CareerArc, a recruitment marketing company
No price was disclosed, and nine of CareerArc's 13 staff are joining Humanly.
- 05
White Circle publishes preprint exposing image model safety flaws
RISE attack achieved explicit-image bypass rates of 13 percent on DALL·E 3, under 10 percent on Nano Banana 2 and about 3 percent on GPT-Image-2; group plans to open-source pipeline.
- 06
Nvidia’s chief executive, Jensen Huang, said he would stop supplying a customer building unsafe AI, speaking to Bloomberg in London
Nvidia’s chief executive, Jensen Huang, said he would stop supplying a customer building unsafe AI, speaking to Bloomberg in London. He described it as a hypothetical but affirmed it, adding that Nvidia would first talk to the customer because its technology is general-purpose. Nvidia has already tightened vetting, removing more than half of its prior customers in Singapore, Malaysia and Japan from its approved-buyer list last year, TNW reported. Huang said he would halt chip supply to a customer building unsafe AI, and his prescription is to accelerate containment and monitoring technology, which Nvidia is building a team to sell. The company has also begun building a safety team to vet agents before deployment, on the logic that trust limits adoption more than capability.
- 07
Salience Labs raises $125M Series B led by Bullhound Capital
The Oxfordshire photonics company will use the money to scale manufacturing and open a second West Coast office.
- 08
Meta launches Small Business Growth Academy across Asia-Pacific
Meta says 84% of owners in a YouGov study of almost 4,000 small business owners currently use AI.
- 09
Forgepoint raises $9.8 million in seed funding
Primary Venture Partners and a16z Speedrun co-led the round for Forgepoint's Frank, an AI agent that produces price quotes.
- 10
Tami investigates findings from CNAPPs and other cloud-native sources, assesses whether a fix is safe
Amazon Web Services says the approach drove an over 90 percent reduction in critical cloud exposure in 90 days.
- 11
Swedish consortium launches Sferical One shared supercomputer
NVIDIA's blog said a Swedish consortium has launched Sferical One, a shared supercomputer running on two NVIDIA DGX SuperPODs with 16 racks of NVIDIA DGX GB300 systems and 1,152 GPUs. The consortium is anchored by Ericsson, AstraZeneca, Saab, SEB and Wallenberg Investments, with SKF as strategic partner. The consortium will build services on NVIDIA Nemotron open models, which the blog says let researchers customize and deploy their models in a secure environment. The post also said AI clouds are expanding in the region, with a projected 1 gigawatt of capacity to be deployed across the Nordics by the end of 2027.
- 12
Ricursive Intelligence aims to automate chip design with AI
Its co-founders said the AI can design chips, learn from each iteration and improve later designs.
- 13
Early access opens for experimental Ask Google Earth deep research
Google said its Geospatial Reasoning agent is an experimental system for multi-step analysis of planet-scale data.