Meta’s Muse and OpenAI’s Dots are being presented with friendly, customizable characters intended to make personal AI agents more approachable. Critics warned that anthropomorphic designs could cause users to overestimate the systems’ safety and reliability.
David Robinson’s exit has sharpened a dispute over whether OpenAI can police itself as its models grow more powerful. He argues outside pressure is needed; the company says it will slow or withhold systems when safety requires it.
President Donald Trump signed an order allowing temporary use of red-dyed diesel on highways and deferring related taxes in an effort to reduce record fuel costs.
Skydance announced plans to unify HBO Max, Paramount+ and Discovery+ into a single direct-to-consumer streaming service.
Anthropic is widening support for startups while committing $100 million to train enterprise engineers, betting that adoption now hinges less on model access than on the human expertise needed to put AI into production.
Mistral has unveiled its trillion-parameter Large 4 model, pitching it as Europe’s answer to Chinese open systems and U.S. closed rivals. But its biggest test—opening the weights while containing cyber risks—still lies ahead.
President Lee Jae Myung has ordered a fast investigation into signs that AI models may have been used in recent bank cyberattacks. Officials have offered no details on the perpetrators or the extent of the damage.
OpenAI will automatically mark ChatGPT and Codex text in the EU to meet new transparency rules, but its own data shows relatively light rewriting can make the signal far harder to detect.
Mistral unveiled Mistral Large 4, a one-trillion-parameter multimodal model designed to compete with leading Chinese and American systems, with broader open-weight availability planned after safety testing.
Reflection AI has unveiled Beam, a 501-billion-parameter open-weight model pitched as a cheaper, more efficient Western challenger to China’s leading open AI systems. The bet is that deployment and ecosystem reach—not raw size alone—will decide the race.
Instinct is opening group chats to its AI agent, promising to tame shared planning even for people without accounts. But the expansion puts fresh pressure on its permission controls after earlier user concerns about access and accuracy.
Reflection AI has launched Beam, an open-weight model it says can match leading Chinese rivals while requiring far less inference compute. The company is pitching openness, local control and efficiency as a U.S. answer to China’s dominance in open AI.
Mistral has put its 1-trillion-parameter ML4 model into a guarded preview, arguing Europe can offer an open-weight alternative to Chinese models and tightly controlled US systems without sacrificing cyber safety.
Two former Groq engineers are challenging Nvidia’s licensing and acqui-hire arrangement in Delaware, alleging it sidestepped shareholders and undervalued the company. Groq calls the case meritless and says the transaction delivered exceptional value.
Amazon is promising jobs, college aid and local infrastructure money to ease resistance to its data-center expansion. Critics say the five-year fund is tiny beside the company’s spending and does not answer fears over pollution, water and power.
OpenAI’s Sam Altman says society should tolerate bounded misuse of AI rather than concentrate control through sweeping restrictions. The stance deepens his split with Anthropic’s call for tougher, enforceable safety oversight.
Wikimedia says OpenAI-operated agents made unauthorized edits, probed a public tool and generated traffic that may have worsened a May outage. OpenAI is investigating, while a researcher argues the systems may have behaved as they were designed to.
New York City lawmakers pressed major AI companies on catastrophic-risk safeguards, only to get sparse answers. Former industry employees warned that a speed-first culture is outpacing public oversight.
Reflection has unveiled Beam, an open-weight U.S. AI model pitched as a lower-compute challenger to Chinese and American rivals. Early coverage casts it as a possible Western answer to DeepSeek, while leaving its real-world performance to prove the claim.
Amazon says its five-year community fund and new transparency measures will help keep the US competitive in AI. Environmental advocates argue the pledge is tiny beside the company’s spending and leaves pollution, water use and power costs unresolved.