The week’s AI news, one entry per event. Stories rank higher when more independent outlets report them, when they are squarely about AI, and when they are recent.
Nick Clegg dismissed fears that AI could gain a 'godlike power to exterminate humanity,' saying many in the tech sector are winding themselves up into a lather and breathing their own fumes.
Amazon introduced new tools for independent sellers on its e-commerce platform, including a plugin that lets them run their businesses through AI assistants.
OpenAI's Astra model passed a driving test, completing it in a parking lot at low speed and at considerable expense.
Google released two text-to-speech models, Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, through its cloud platform. The models have highly similar application programming interfaces, which makes using them side-by-side relatively easy.
Apple researchers describe a method for compressing streaming neural audio encoders using latent-space distillation, aimed at the on-device dictation pipeline that feeds a tokenizer encoder into a foundation model.
Apple researchers describe a semi-supervised federated learning recipe for automatic speech recognition that uses a teacher to generate pseudo-labels on clients' unlabeled data, with a small labeled seed set on the server, plus server update stabilization.
Meta said it is bringing its Muse AI agent to its smart glasses, including glasses unveiled at Meta Connect, and announced updates such as giving each agent its own email address and adding video chat support.
A safety researcher published a post reflecting on the distinction between "safety" and "capabilities" work in AI, written as part of the MATS 9.1 extension program under mentor Richard Ngo.
The US rejected requests from OpenAI, Anthropic and Hugging Face for global AI standards, according to the BBC. Sam Altman, Dario Amodei and Clement Delangue had each called for risk evaluation standards.
ChatGPT Voice now runs on OpenAI's GPT-6 Astra, Sol, and Luna models and can access plugins including email, calendar, and Slack, letting users manage appointments, send emails, or build websites by talking.
OpenAI set pricing for models called GPT-6 Sol and Luna, a move described as escalating an AI price war in response to Google, Anthropic and open source models.
Rabbit Inc. unveiled OS3, a cloud-based personal AI agent that connects to personal devices.
Microsoft Research published findings showing that moving AI inference off a robot's onboard hardware can improve task success, boost efficiency and support more advanced physical AI workloads.
The Verge's Optimizer newsletter discusses Meta's Muse AI mascot, calling it sinister that the mascot is so cute. The author describes asking their own renamed version of the mascot a question.
A guest post argues that in science, thinking has become cheap while doing remains expensive, and that this asymmetry is quietly changing how research companies operate.
The Register is hosting an off-the-record roundtable with HPE and NVIDIA in Washington, chaired by Matt Rosoff, on data sovereignty in the AI era.
A post on LessWrong offers loose thoughts on the persona selection model, arguing it is over-applied as AIs have become more "RLVR-brained."
A post describes Jev, a new model format that outputs certainties over a defined set of options instead of text, and applies it to monitoring harmful thought traces.
The US rejected requests from OpenAI, Anthropic and Hugging Face for global AI standards.
Experts say the Trump administration's focus on winning an "AI race" against China may discourage China from sharing safety intelligence.
Basecamp Research Ltd. raised $140 million in a Series C round led by S32, a fund affiliated with Bill Maris.
Enveda raised $311 million, valuing the AI biotech at $2 billion. It is currently testing drugs for skin conditions and for preserving weight loss after stopping GLP-1s.
Skild AI, a company backed by Nvidia, has developed a method that lets robots learn new tasks from a single video. It is aimed at general-purpose robots and real-world deployment of physical AI.
Paolo Benanti told WIRED that concern over whether AI could destroy humanity is distracting from the need for public debate about governing the technology. He also expressed worry about 'cartel' behavior among big labs.
Ema Unlimited Inc. raised $77 million in a Series B round led by Creagis to expand its autonomous AI workers into enterprise HR, IT and finance departments.
Alibaba's Qwen team released Qwen-Audio-3.1, a lineup of five models for speech recognition, text-to-speech, and real-time interaction, and cut AI audio prices by up to 95 percent. The ASR model improves multilingual and dialect recognition and automatically cleans up filler words.
Donald Trump welcomed Xi Jinping to the White House for a summit that included talks on AI, trade, Taiwan and Iran. The meeting was heavy on symbolism, with no sign that closed-door discussions produced breakthroughs on those issues.
A researcher on LessWrong describes having low confidence in findings about AI use in research and publishing, saying AI usage in research has been changing quickly in recent months.
Australia announced a breach of OpenAI at the United Nations.
Flock's CEO declined to appear at a Senate hearing Wednesday about the company's "AI Surveillance Network," where senators criticized the company.