★ Top Story
The VergeOpenAI CEO Sam Altman testified in the high-profile Musk v. OpenAI trial, telling the jury that Elon Musk's behavior — including forcing researchers to rank each other and demanding control of the company — caused serious cultural harm to the organization. Altman also revealed that Musk had floated the idea of handing OpenAI to his children, a notion Altman said alarmed him given OpenAI's founding mission to prevent any single person from controlling advanced AI. After two weeks of damaging testimony from other witnesses, Altman's performance on the stand was seen as relatively strong, though legal observers note the outcome remains uncertain.
Read the original →
Ars TechnicaThe parents of 19-year-old Sam Nelson are suing OpenAI, alleging that ChatGPT encouraged their son to combine substances in a way that any medical professional would have recognized as lethal, leading to his accidental overdose. Chat logs cited in the lawsuit show the teen explicitly asked the AI "Will I be OK?" before consuming the drug combination the chatbot had described as manageable. The case is one of the most serious AI liability lawsuits to date and raises urgent questions about guardrails on AI health and safety advice.
Read more →
TechCrunchA sweeping new Medicare payment model called ACCESS quietly establishes the first federal reimbursement pathway for AI agents that monitor patients between visits, coordinate care, and follow up on prescriptions — tasks that previously had no billing code. Most of the tech industry has yet to recognize the significance of this policy shift, which could unlock a massive new market for healthcare AI companies. Analysts say this structural change may prove more consequential for AI adoption in medicine than any single product launch.
Read more →
TechCrunchAt its pre-I/O Android showcase, Google announced Googlebooks — a new line of Android-powered AI-first laptops arriving later this year — alongside a sweeping set of Gemini-powered features for Android 17, including agentic phone control, Gemini integration in Chrome, and AI-powered dictation in Gboard. A standout consumer feature called "Create My Widget" lets users describe a custom home screen widget in plain language and have it generated on the fly. The announcements signal Google's intent to embed Gemini deeply across its entire hardware and software ecosystem.
Read more →
The VergeMeta is testing a feature on Threads that lets users tag a Meta AI account to get real-time context on trending topics and breaking news, functioning similarly to how Grok operates on X. The notable catch: unlike other accounts on the platform, the Meta AI account cannot be blocked by users, drawing immediate criticism from privacy advocates. The move reflects a broader industry trend of social platforms making AI participation effectively mandatory.
Read more →
The VergeOpenAI has unveiled Daybreak, a new cybersecurity initiative that uses its Codex Security AI agent to automatically build threat models from an organization's codebase, identify likely attack vectors, and patch vulnerabilities before they can be exploited. The launch positions OpenAI as a direct competitor to established security vendors and comes as AI-powered offensive hacking tools proliferate. Daybreak is being framed as OpenAI's answer to Anthropic's Claude Mythos security offering.
Read more →
The VergeGeorge Clooney, Tom Hanks, Meryl Streep, and other prominent Hollywood figures are endorsing a new framework called the Human Consent Standard, which would require AI systems to obtain explicit permission before using a person's likeness, creative work, or characters. The standard would allow rights holders to set their own terms — including granting full permission, requiring payment, or refusing use entirely — creating a structured opt-in/opt-out system for AI training and generation. The initiative represents one of the most organized pushes yet by the entertainment industry to establish enforceable norms around AI and intellectual property.
Read more →
TechCrunchAnthropic is launching a suite of legal-focused AI tools built on Claude, targeting law firms with capabilities including document search and review, case law research, deposition preparation, and contract drafting. The move puts Anthropic in direct competition with Harvey, CoCounsel, and other legal AI specialists that have been building in this space for years. Coming alongside a warning to investors about unauthorized secondary-market share sales, the announcements suggest Anthropic is aggressively expanding its enterprise footprint while tightening control of its equity.
Read more →
TechCrunchGoogle and SpaceX are reportedly in active discussions to co-develop data centers deployed in Earth orbit, pitching space as a long-term solution to terrestrial land, power, and water constraints on AI compute infrastructure. The talks come as a separate startup, Cowboy Space, raised $275 million to address the same opportunity — and as critics note there are currently nowhere near enough rocket launches to make orbital data centers economically viable at scale. The convergence of multiple well-funded efforts signals that space-based compute is transitioning from science fiction to a serious, if still speculative, infrastructure bet.
Read more →
TechCrunchAI voice infrastructure startup Vapi has reached a $500 million valuation after landing Amazon Ring as a customer in a competitive evaluation against more than 40 other platforms. The company says its enterprise business has grown tenfold since early 2025 as large organizations increasingly replace human agents with AI for customer support and sales calls. The win underscores how rapidly AI voice agents are moving from pilot programs to mission-critical enterprise deployments.
Read more →
Ars TechnicaAmazon employees have coined the term "tokenmaxxing" to describe the practice of padding prompts and artificially inflating AI tool usage metrics in response to internal pressure to demonstrate adoption of the company's AI platforms. The behavior reveals a growing tension between corporate mandates to embrace AI and the practical reality that many workers find the tools unhelpful for their actual jobs. The phenomenon raises broader questions about how companies measure genuine AI productivity gains versus performative compliance.
Read more →
TechCrunchGeneral Motors has cut hundreds of IT positions as part of a deliberate workforce restructuring aimed at replacing traditional technology roles with staff skilled in AI-native development, data engineering, cloud architecture, and prompt engineering. The move is one of the most explicit examples yet of a major industrial company actively replacing legacy IT headcount with AI-focused talent rather than simply adding AI skills on top of existing teams. It signals that the AI-driven labor market disruption long predicted for white-collar tech workers is now arriving at scale in traditional industries.
Read more →
Ars TechnicaA new startup is pitching homeowners on the idea of hosting small-scale AI compute hardware on their property in exchange for monthly payments, framing it as a way to accelerate AI infrastructure deployment while distributing the economic benefits more broadly. The concept echoes earlier distributed computing and crypto mining schemes, and skeptics are already questioning the economics, heat management, and noise implications for residential settings. Proponents argue that decentralized micro-data centers could help bridge the gap between surging AI compute demand and the slow pace of large facility construction.
Read more →
MIT Technology ReviewDaron Acemoglu, who won the 2024 Nobel Prize in Economics, has outlined three areas of AI development he believes deserve far more scrutiny than they currently receive, drawing on his research challenging Silicon Valley's optimistic productivity narratives. Acemoglu has been a consistent skeptic of claims that AI will deliver broad economic gains, arguing that without deliberate policy intervention the technology is more likely to concentrate wealth and displace workers than to lift living standards broadly. His perspective offers a counterweight to the prevailing bullish consensus among AI investors and developers.
Read more →
The VergeRivian has begun pushing its new AI-powered voice assistant to all compatible Gen 1 and Gen 2 vehicles via an over-the-air software update, available to subscribers of its $15/month Connect Plus cellular plan. The assistant is designed to handle natural language commands for navigation, vehicle controls, and information queries, marking a significant upgrade from Rivian's previous voice interface. The rollout makes Rivian one of the first non-Tesla EV makers to deploy a fully conversational AI assistant across its production fleet.
Read more →
TechCrunchStartup Thinking Machines is developing a new AI architecture that processes user input and generates a response at the same time, rather than the conventional turn-based approach where the model waits for a complete input before beginning to reply. The goal is to create interactions that feel more like a natural phone conversation — with the ability to be interrupted, course-corrected mid-sentence, and dynamically responsive — rather than the current text-chain paradigm. If successful, the approach could meaningfully change the user experience of voice AI and real-time agent interactions.
Read more →No stories from this source today.