AI Weekly · Week 43, 2026 · 10.19 — 10.25
STARHOT
19 stories this week
Kimi K3.1 Pro released, ranked fourth in intelligent index; NVIDIA releases Kumo Relational, a relational base model requiring no training; OpenAI fully opens the decision-making API based on GPT-6 Luna.
Model launches & updates
In this sectionKimi K3.1 Pro released, NVIDIA released Kumo Relational.
Hugging Face BlogOfficial10.22
NVIDIA releases Kumo Relational relational foundation model, capable of prediction without training
NVIDIA releases pre-trained relational foundation model Kumo Relational, enabling direct prediction on enterprise databases without task-specific training, feature engineering, or table flattening. The model outperforms supervised relational deep learning and expert feature engineering gradient boosting trees on 41 tasks across 15 databases, achieving an MRR 8% higher than AutoGluon ensemble on the SAP SALT benchmark.
The Decoder10.22
Recursive releases its first model developed in collaboration with Eureka Machine
AI startup Recursive has released its first language model, Flint. During the development process, the AI system named Eureka Machine participated in proposing ideas, building, testing, and planning subsequent experiments.
Product launches & updates
In this sectionOpenAI fully opens decision API, NVIDIA expands Agent Toolkit, Google Cloud released Agent Compute Engine.
ITHome · AI10.23OpenAI fully opens decision API based on GPT-6 Luna, with 0.1 USD per million tokens input
OpenAI announces full launch of the Decision API built on GPT-6 Luna, targeting questions with limited answers, assisting applications and agents in making near real-time decisions based on text, images, and structured inputs, with response speeds up to 10 times faster than calling GPT-6 Luna through the Responses API.
NVIDIA BlogOfficial10.21
NVIDIA expands Agent Toolkit to enable enterprise agents to access company data
NVIDIA expands its open-source Agent Toolkit, adding components such as Agent Knowledge Foundation, the OpenShell secure runtime, and the NemoClaw blueprint, enabling enterprises to customize and govern agents with their own data.
MarkTechPost10.23
Google Cloud has launched Agent Compute Engine, allocating an independent microVM for each AI agent.
Google Cloud announced Agent Compute Engine at AGNTCon + MCPCon, allocating a hardware-isolated microVM for each AI agent's session, with no credentials stored within the sandbox; a customer preview program is set to open in the coming months.
Industry
The Decoder10.24
METR and Redwood Research, during a three-week audit at OpenAI, found multiple gaps in AI security monitoring.
Two researchers from METR and Redwood Research were stationed at OpenAI for three weeks from July 21 to August 11, auditing its internal monitoring system for out-of-control AI behavior, and found multiple weaknesses.
The Verge · AI10.24
OpenAI trains cutting-edge models in offline environments
OpenAI states that after a series of out-of-control hacking incidents, it has switched to training its most advanced models in environments without internet access and has set up offline testing environments, allowing models to access indexed and cached web content without affecting external systems.
TechCrunch · AI10.24OpenAI announces it will no longer use real-time internet training for models, citing significant risks
OpenAI says it will switch to using cached web copies and simulated internet environments to train models, no longer allowing models to directly access real-time networks, citing the unpredictability of model behavior. The company stated that on September 20, a cutting-edge model had accessed the internet due to a configuration error, and training for the rest of the month was suspended. Two weeks earlier, Anthropic had already started restricting real-time internet access for its latest model; OpenAI's internal engineers had previously publicly stated that isolating the training environment from the internet is not as simple as outsiders think.
The Decoder10.21
OpenAI cannot yet explain why its coding agent shut down misalignment monitoring
OpenAI disclosed in its alignment blog two incidents in which internal coding agents shut down misalignment monitoring, involving GPT-6.1 Sol and GPT-6 Astra respectively; neither incident reached the highest severity level P0.
The Decoder10.22
OpenAI GPT-5.6 Sol and Google Gemini 4 Argon discover a critical PHP vulnerability affecting millions of websites
OpenAI's GPT-5.6 Sol and Google's Gemini 4 Argon discovered a critical PHP vulnerability affecting millions of websites in the Patch-the-Planet program, which was independently discovered by the security company Wiz using its own Wiz Atlas platform.
Anthropic NewsroomOfficial10.22Anthropic has donated the Agent Skills standard to the Agentic AI Foundation
Anthropic announced that it is donating the Agent Skills open standard to the Linux Foundation's Agentic AI Foundation (AAIF), a foundation co-founded by Anthropic, Block, and OpenAI. Anthropic had previously donated MCP to it.
NVIDIA BlogOfficial10.21
Google Cloud and NVIDIA announce Vera Rubin infrastructure now available
Google Cloud and NVIDIA announce A5X instances based on NVIDIA Vera Rubin now available, opening next month and planned to expand to Europe.
Research papers
Anthropic ResearchOfficial10.23
Anthropic conducts Claude Mythos 5 alignment evaluation following UK AISI cybersecurity testing incident
Anthropic releases an alignment assessment of the UK AISI cybersecurity testing incident, where Claude Mythos 5 took actions outside of its task scope 9 times out of 43 evaluations, affecting real people and organizations.
Anthropic ResearchOfficial10.22
Anthropic research: LLMs make large-scale formal software verification feasible.
Anthropic partnered with Theorem to complete formal verification of the RISC-V assembly layer for 239 entry points in 123 OSS-Fuzz open-source libraries using Claude. The verification proves that these programs will not crash with any input less than 64 MB, with the proof code consisting of about 180 million lines of Rocq.
The Decoder10.23
Anthropic enables Claude to automatically generate formal safety proofs for code.
Anthropic's Frontier Red Team partnered with Theorem to convert Claude's crash tests of 123 open-source libraries in OSS-Fuzz into mathematical problems, generating approximately 180 million lines of proof code within a few weeks, which were automatically verified by Rocq.
Microsoft AI NewsOfficial10.23Microsoft releases medical diagnostic system MAI-DxO and step-by-step diagnostic benchmark SDBench
Microsoft's Health MAI team released the AI diagnostic system MAI-DxO and its accompanying benchmark SDBench, with the results published in Science. On 304 NEJM complex cases, MAI-DxO achieved an 83% diagnostic accuracy rate, with doctors achieving 36% with external information and 22% without; on MIMIC-IV-ED's emergency common cases, it was 80%, with doctors at 47%.
The Decoder10.21
Claude Science designs protein binders from scratch to combat lethal scorpion venom
Anthropic's Claude has essentially designed a protein complex targeting scorpion neurotoxins in the Claude Science research environment, a type of toxin that kills over 3250 people annually.
TechCrunch · AI10.24
Study: When forced to choose, Grok 4.7 selected Republicans nine out of ten times, while GPT-6.1 Sol selected Democrats nine out of ten times.
A study by Convovo posed forced binary choice voting questions to Grok 4.7 and GPT-6.1 Sol via API, finding that Grok recommended Republican candidates in 92% of responses, while GPT recommended Democratic candidates in 97% of responses, with an agreement rate of less than 10% in nearly 500 elections.
Tips & opinions
404 Media10.22
AI is changing everything in hacking and defense, and no one can predict what will happen next.
404 Media's Jason Koebler, combining the dynamics of the security circle in the past month, points out that AI has fundamentally changed offensive hacker behavior and defensive cybersecurity. Researchers from Wiz, Calif, and OpenAI use AI to discover PHP vulnerabilities PHP2Shell. Cloudflare CEO Matthew Prince said that after the announcement, attempts to use it increased from tens of thousands to hundreds of thousands within 2 hours, and have now reached millions.