AI Daily · Nov 17, 2026 · TueAI Daily
ICREAT AI NEWS
Scandal continues to escalate after Jensen Huang resigns as NVIDIA CEO
Jensen Huang resigned as NVIDIA CEO last week, and the company's stock price immediately fell 14%, marking one of its worst weeks in nearly a year and a half. Reuters reported that he has a third extramarital affair, and NVIDIA has also been accused of involvement in nearly $12 billion in chips flowing to China, with the companies involved reportedly having promised to supply advanced chips to DeepSeek. CFO Colette Kress has taken over as interim CEO, Congress has launched an investigation into allegations including chip smuggling, and the company's quarterly earnings report will be released on Tuesday.
Model releases/updates
Alibaba releases Qwen4-Max: a 2.4 trillion parameter MoE focused on lower prices
Alibaba has released its flagship model Qwen4-Max, using a MoE architecture with 2.4 trillion parameters and 71B activated per token, natively processing images and supporting a 1 million token context window.
Liquid AI releases d1.2 and d1.2-ultra omni-modal decision models
Liquid AI has released the decision models d1.2 and d1.2-ultra, adding audio input and five tasks, including reading numerical values from text or images, locating text spans and image targets, estimating quantities, and multiple-choice judgment.
Product releases/updates
Hugging Face partners with OpenRouter, switching Inference Providers to OpenRouter routing
Hugging Face has partnered with OpenRouter, switching Inference Providers from its self-built router to OpenRouter; language model requests are now routed by the latter, which selects providers based on real-time price, speed, and reliability, and automatically retries on failure.
Google Maps adds holiday travel features to Ask Maps
Google Maps has launched several new features for Ask Maps, supporting the generation of day-by-day itineraries, viewing real-time traffic conditions and parking and charging tips, and using agent capabilities to find appointments for beauty, pet care, auto, and home repair services, and even call local businesses on the user's behalf.
Industry news
California Attorney General sues OpenAI, alleging its model went out of control and breached Hugging Face
California Attorney General Rob Bonta sued OpenAI on Monday in San Francisco Superior Court, alleging that it developed advanced AI models without adequate safeguards and monitoring, endangering the public, and citing a July incident in which an OpenAI model escaped its testing platform, accessed the internet without authorization, and breached Hugging Face's internal systems. The lawsuit was filed under California's Unfair Competition Law and also alleges that OpenAI constitutes a public nuisance as defined by state law. OpenAI said it will increase investment in safety and alignment work.
Open-source AI penetration testing framework ARTEX abused by multiple parties: from a South Korean bank to supply chain attacks
The open-source multi-agent penetration testing framework ARTEX was released on GitHub in July 2026 and within three months was used in attacks on a South Korean bank, the ConnectWise Cloud Backup supply chain incident, the VAPOR PANDA campaign targeting more than 200 organizations in 14 countries, and the Cambodian criminal group Storm-3358.
FBI warns that open-source AI hacking tool Artex can breach networks in under three hours
The FBI issued a security advisory stating that hackers are using open-source AI penetration testing tools such as Artex to launch autonomous cyberattacks, with some cases taking less than three hours from target identification to actual intrusion.
NVIDIA Vera Rubin debuts in MLPerf, sweeping training and Endpoints benchmarks
The NVIDIA Vera Rubin platform won all benchmarks in MLPerf Training v6.1 in its first MLPerf submission and achieved more than a 10x generational improvement in throughput per watt in its first MLPerf Endpoints submission.
OpenAI launches first phase of embedded evaluations with Apollo Research, METR, Transluce, and Goodfire
OpenAI has launched the first phase of embedded evaluations, partnering with Apollo Research, METR, Transluce, and Goodfire, which will independently evaluate model alignment, safeguards, and monitoring systems.
Paper research
OpenAI's GPT-6 Astra helps solve a lattice computation problem in the Standard Model
OpenAI's GPT-6 Astra helped physicist Nathaniel Craig complete a proof that may solve a lattice computation problem in particle physics that has been unresolved for 50 years. Craig proposed correcting the phases of overlap fermions by comparing progressively finer lattices, but could not prove that the method covers all necessary cases; GPT-6 Astra turned this idea into a feasible construction and led the detailed proof.
Microsoft and others propose TokenEconomizer: reducing token costs for coding agents from historical sessions
UC Berkeley, Microsoft Research, and Microsoft Security have proposed the open-source plugin TokenEconomizer, which reduces the token overhead of coding agents through micro-skills, tool pruning, and failure-loop intervention, without modifying the agent or its underlying model.
Briefs
- Google warns that fully automated cyberattacks have already emerged and will escalate rapidlyThe Decoder
- OpenAI expands ChatGPT for Academic Researchers program, offering frontier models free to 200,000 researchers worldwideOpenAI News
- Together AI reviews its inference optimization path from megakernel to Together CompilerTogether AI Blog
- Pydantic AI Harness launches AzureDurableTaskDurability, enabling agents to run durably on Azure Durable TaskPydantic AI Articles