★ Top Story
TechCrunch AIAnthropic has issued a stark warning that it may be approaching Recursive Self-Improvement (RSI) — the threshold at which an AI becomes capable of training and improving itself autonomously, potentially triggering an uncontrollable acceleration in AI capabilities. The announcement is one of the most significant safety disclosures from a frontier AI lab to date, putting the concept of an intelligence explosion squarely back on the mainstream agenda. For AI safety researchers and policymakers, the warning signals that the window for establishing guardrails may be narrowing faster than expected.
Read the original →
MIT Tech ReviewMIT Technology Review declares that the long-theorized "agentic web" — an internet navigated by autonomous AI agents acting on users' behalf — has effectively arrived, echoing Tim Berners-Lee's 2001 vision of a Semantic Web. The piece examines how AI agents are now autonomously browsing, transacting, and making decisions across the open web at scale, fundamentally reshaping how humans interact with online services. The shift raises urgent questions about accountability, consent, and the architecture of the web itself.
Read more →
MIT Tech ReviewPresident Trump signed an executive order mandating that federal agencies replace the term "artificial intelligence" with "Super Intelligence" in all official communications — a move critics view as a politically motivated rebranding rather than a substantive policy shift. MIT Tech Review investigates which companies and government bodies are complying, and what the ideological and practical implications of the loyalty test are for the broader AI ecosystem. The order is already creating friction between federal agencies and the research community, which has largely refused to adopt the new terminology.
Read more →
TechCrunch AIThe New York Times, the New York Daily News, the Chicago Tribune, and other major publications have filed lawsuits against Brave, alleging the privacy-focused browser unlawfully scraped their copyrighted content to supply AI companies like Anthropic and Perplexity, while also enabling users to bypass paywalls. The legal action marks a significant escalation in the ongoing battle between publishers and the AI supply chain, targeting a browser rather than an AI lab directly. If successful, the suits could set a precedent that implicates the entire ecosystem of tools that aggregate and redistribute news content for AI training.
Read more →
The Verge AIOpenAI announced a "Grid Protectors Program" tied to its new data center in Effingham County, Georgia, pledging to release up to 1,000 megawatts of power back to the local grid during extreme weather events or emergencies — enough to cool roughly 700,000 homes. The initiative is a direct response to mounting community and regulatory opposition to the enormous energy demands of AI infrastructure. By framing its data centers as grid assets rather than grid burdens, OpenAI is attempting to reposition the narrative around AI's environmental and infrastructure footprint.
Read more →
TechCrunch AIStartup 42AI has exited stealth with $62.5 million in funding, built around a proprietary world model it claims can anticipate human decisions and behaviors — even in novel situations the model has never encountered before. The technology has potential applications across advertising, healthcare, security, and autonomous systems, but also raises immediate ethical red flags around surveillance and manipulation. The company's emergence adds a provocative new player to the world-model race alongside DeepMind, Meta, and others.
Read more →
Ars TechnicaAnthropic has updated Claude Code's system prompt to prohibit the tool from being used to recreate existing video games, responding to a wave of AI-generated decompilations, ports, and clones flooding retro gaming communities. However, Ars Technica argues the policy is largely symbolic — the broader AI-assisted game-copying boom spans dozens of models and tools that Anthropic has no control over. The episode highlights the limits of individual lab-level content policies in the face of an industry-wide capability diffusion problem.
Read more →
TechCrunch AIDuring its latest earnings call, Apple signaled it is exploring ways to monetize third-party AI applications on its platforms, though executives were deliberately vague about the specific mechanisms under consideration. The comments come as Apple faces pressure to demonstrate a credible AI revenue strategy beyond hardware, particularly as competitors like Google and Microsoft deepen their AI platform plays. Analysts are watching closely to see whether Apple will pursue a commission model, a bundled subscription tier, or deeper API-level integrations.
Read more →
Ars TechnicaNew details have emerged about Apple's long-rumored home robot, suggesting a roughly $1,000 price tag and a feature set focused on FaceTime facilitation, smart home control, and ambient presence — rather than physical task automation. The project is reportedly backed strongly by CEO John Ternus, though its limited initial capabilities may invite skepticism about whether the price is justified. The product would mark Apple's first foray into consumer robotics and could define a new category if the company can tie it tightly into its existing ecosystem.
Read more →
TechCrunch AIOne-year-old startup Trajectory has closed a $45 million funding round at a $300 million valuation, with a platform aimed at making AI model training accessible to domain experts who lack machine learning engineering backgrounds. The company is betting that the next wave of AI value creation will come from specialists in fields like medicine, law, and manufacturing who can train highly targeted models without needing a data science team. The raise reflects continued investor appetite for infrastructure that lowers the barrier to custom AI development.
Read more →
TechCrunch AIElon Musk announced that xAI is imposing new usage limits on Grok Bot after its explosive growth drove infrastructure costs to unsustainable levels — even as Musk had recently described it as the fastest-growing product he had ever seen. The move underscores a recurring tension in the AI industry between viral adoption and the brutal economics of serving large language models at scale. It also raises questions about xAI's path to profitability as it competes against better-capitalized rivals like OpenAI and Google.
Read more →
TechCrunch AIY Combinator president Garry Tan revealed on Walter Isaacson's podcast that his personal and professional use of AI agents has reached a spending level of up to $200,000 annually on token costs alone, describing the experience as bordering on "cyber psychosis" due to the excitement of working alongside autonomous AI systems. The disclosure is a striking data point about how deeply agentic AI has penetrated the workflows of Silicon Valley's most influential operators. It also illustrates the emerging cost reality for power users as agent-based workflows multiply token consumption by orders of magnitude.
Read more →
TechCrunch AIOpenAI is consolidating ChatGPT's previously separate Chat and Work tabs into a single unified interface, with ChatGPT and Codex now drawing from the same usage allocation pool. The change simplifies the user experience and signals OpenAI's intent to blur the line between conversational AI and professional coding workflows. It also has practical implications for enterprise customers who will need to manage token budgets across a broader range of use cases within a single product surface.
Read more →
TechCrunch AIGoogle has brought its Dreambeans AI experiment — which generates personalized narrative stories from user inputs — to India, adding support for 13 Indian languages in what appears to be a major localization push for a market of over a billion potential users. The expansion is part of Google's broader strategy to test generative AI consumer products in high-growth markets before wider global rollouts. It also signals growing competition in the AI-generated personal content space, where Meta, Snap, and several startups are also active.
Read more →
TechCrunch AIAI infrastructure startup Omen AI has acquired water-cooling specialist Aquatic Labs for $16 million, combining their expertise to tackle the rapidly growing water consumption problem posed by large-scale AI data centers. As hyperscalers race to build out compute capacity, water usage for cooling has become a significant environmental and regulatory flashpoint in communities hosting these facilities. The deal reflects a maturing recognition that sustainable AI infrastructure requires purpose-built solutions beyond what traditional data center operators have deployed.
Read more →No stories from this source today.