News

Google announced a three‑year agreement to rent roughly half of SpaceX’s AI‑compute capacity, paying $920 million per month from October 2026 through June 2029. The deal, disclosed in a regulatory filing, gives Google access to about 110,000 NVIDIA GPUs and associated hardware. It mirrors SpaceX’s earlier contract with Anthropic, though at a lower scale, and comes just weeks before SpaceX’s anticipated IPO. Both companies say the partnership addresses surging demand for Google’s Gemini Enterprise AI platform, while a cancellation clause lets either side walk away with 90‑day notice after the end of 2026. Read more

Airbnb chief executive Brian Chesky said Wednesday he will fund a new artificial‑intelligence research lab that will concentrate on visual and conversational interfaces, while remaining at the helm of the home‑rental platform. The move positions the Silicon Valley veteran against OpenAI, whose CEO Sam Altman – a longtime associate of Chesky – recently returned to power after a board revolt. Chesky’s plan reflects growing frustration with existing large‑language‑model products, which he says fall short of the rich, visual experiences needed for travel and commerce. Read more

Token costs have dropped 98%, yet enterprise AI bills have surged by more than threefold as agentic tools drive unprecedented consumption. Companies like Uber and Priceline have already exceeded their 2026 AI budgets, while Microsoft pulled developer licences after costly overspend. In response, the Linux Foundation announced the Tokenomics Foundation, a standards body aimed at bringing cost discipline to AI token usage. Industry leaders warn that without guardrails, trillions of token transactions could overwhelm finance teams. Read more

About 90 venture‑capital firms have stakes in both OpenAI and Anthropic, the two leading AI labs that have been competing for talent, customers and policy influence. Data from PitchBook shows roughly 42% of OpenAI’s investors also back Anthropic, and a third of Anthropic’s investors sit on OpenAI’s cap table. The overlap includes heavyweight names such as Sequoia Capital, Greylock, Founders Fund and Amazon. With both companies planning public listings this year after raising over $100 billion each, investors appear to be betting on the broader AI market rather than picking a single winner. Read more

Anthropic, one of the leading firms in the artificial‑intelligence race, warned that the speed of AI progress could soon enable systems to design their own successors. In a recent blog post, the company urged a temporary worldwide slowdown to give policymakers and researchers time to develop safety and alignment measures. While Anthropic touts its upcoming profitability and a pending public offering, critics argue the pause plea doubles as a marketing strategy, pointing to the limited rollout of its cybersecurity model Mythos. The firm plans to convene industry and government leaders in the coming months to explore how a coordinated pause might work. Read more

OpenAI announced it will comply with President Donald Trump’s executive order that calls for government review of advanced artificial‑intelligence models before they reach the public. The company said it will submit its frontier models for assessment, citing the need for democratic oversight while urging regulators to build flexible, powerful bodies. The order, originally drafted with industry input and a 90‑day review window, was trimmed to a 30‑day voluntary benchmark after pushback from tech leaders. Critics argue the scaled‑back measures fall short of ensuring safety. Read more

Anthropic reports that its AI system Claude authored more than 80% of the company’s production code as of May 2026, a jump from single‑digit levels a year earlier. Engineers now ship eight times more code per quarter, and the model’s success rate on complex problems has risen to 76%. The firm’s new Anthropic Institute paper calls for a verifiable, multinational pause mechanism on frontier AI work, likening it to nuclear arms‑control agreements. Anthropic says the rapid automation of software development could reshape the industry, but also warns of the risks of unchecked recursive self‑improvement. Read more

Former OpenAI CTO Mira Murati reappeared on a Bloomberg interview Thursday, marking her first major public appearance in roughly 18 months. As CEO of Thinking Machines Lab, she previewed the company’s new “interaction models,” a full‑duplex AI interface that processes continuous audio, text and video streams in 200‑millisecond slices. Murati also highlighted the firm’s $2 billion fundraising, a gigawatt of Nvidia Vera Rubin compute and the departure of several senior researchers. She used the platform to argue that the AI industry lacks structural governance checks, revisiting the 2023 OpenAI board crisis that thrust her into the spotlight. Read more

The National Security Agency has placed roughly six Anthropic engineers on its premises to assist with the deployment of the company’s advanced cybersecurity AI model, Mythos, according to a Financial Times report. While the move signals a deeper partnership between the agency and the AI firm, officials declined to confirm whether the model is already being used in offensive cyber missions. The collaboration follows a federal ban on Anthropic technology after the Department of Defense labeled the company a supply‑chain risk, heightening scrutiny over AI’s role in national security. Read more

Tech giants and startups alike are confronting runaway AI token costs as enterprise budgets blow past projections. Uber exhausted its 2026 AI coding allocation by April, Microsoft pulled back developer licenses, and Priceline saw a routine contract renewal jump fivefold. The spike stems from aggressive adoption of large language models and autonomous agents, which have amplified consumption despite lower per‑token prices. In response, a coalition led by the Linux Foundation is forming the Tokenomics Foundation to create industry‑wide standards for tracking, auditing, and optimizing AI spend, while vendors scramble to offer new cost‑control tools. Read more

A new study published in PNAS Nexus reveals that leading large‑language models, including GPT‑4o and Claude 3.5 Sonnet, perform poorly on the Stroop effect test, especially as task length increases. While humans maintain roughly 95% accuracy even in extended trials, the AI systems’ accuracy drops sharply, highlighting a fundamental gap in executive attention that researchers say must be addressed before artificial general intelligence can be realized. Read more

Meta has yet to release an API for its Muse Spark large‑language model, despite promising developers the tool would be available in June. The Wall Street Journal reported that the company announced the model in April but provided no firm launch date for the API, leaving developers who want to integrate the closed‑source AI into their applications in limbo. A Meta spokesperson reiterated the June target, but weeks have passed without a public rollout. The delay adds pressure on the social‑media giant as rivals such as Microsoft, Google and Apple push their own AI services forward. Read more

Cloudflare CEO Matthew Prince announced Wednesday that AI‑driven agents now generate more internet traffic than human users worldwide. Data from the company's Radar service shows agentic bot activity accounting for 57.4% of total web requests, while human‑originated traffic fell to 42.6%. Prince said the shift happened faster than he expected, noting that the trend is already evident across most regions, with North America seeing the highest bot share. The rise reflects the growing use of AI chatbots that browse the web on users' behalf. Read more

Anthropic, the AI model developer that recently raised $65 billion at a $965 billion valuation, has confidentially filed for an initial public offering. Co‑founder Daniela Amodei told Bloomberg Tech that the move is driven by the massive capital needed to train and run large models. The company reported annualized revenue of more than $47 billion in May, a sharp jump from roughly $9 billion at the end of 2025, underscoring rapid growth even as some firms reassess AI spending. Read more

Mira Murati, former OpenAI CTO and now chief executive of Thinking Machines Lab, gave her first extensive media interview in 18 months. Speaking with Bloomberg in San Francisco, she previewed the startup’s new “interaction models,” AI systems that process continuous streams of audio, text and video in near‑real time. Murati also reflected on her brief stint as interim OpenAI CEO during the 2023 board upheaval and warned that the industry’s decision‑making is too concentrated. The interview signals Thinking Machines’ push to re‑enter the public AI arena while underscoring governance concerns. Read more

Airbnb chief executive Brian Chesky is set to back a new artificial‑intelligence laboratory, a move that signals the hospitality platform’s deeper push into AI research. While Airbnb has already integrated AI‑driven coding tools, Chesky said the company has not yet formed a partnership with any large‑language‑model provider because existing products do not meet its standards. The new lab, whose focus is expected to center on user interaction and design, will operate under Chesky’s oversight but will not be led by him personally. A company spokesperson declined to comment further. Read more

Anthropic and other AI firms have filed to go public, sparking a wave of real‑estate listings that accept company stock as payment. Meanwhile, a former NLRB employee has sued Elon Musk for defamation after the billionaire dismissed his whistleblower allegations about the DOGE hack. At the same time, hackers exploited Instagram’s AI chatbot to hijack high‑profile accounts, prompting Meta to scramble for a fix. Read more

Meta unveiled Creator Assistant, an AI-driven tool embedded in the Facebook dashboard that lets creators ask questions about their performance and get content ideas. Available now to creators in the United States, Canada and India, the conversational assistant pulls data from a user’s own Facebook presence and trending topics to deliver actionable insights. While the feature promises to simplify analytics and spark new ideas, it also requires full account access, raising privacy and security concerns after a recent AI‑based support bot was reportedly hacked. Read more

Prime Minister Mark Carney announced a five‑year "AI for All" strategy that seeks to boost Canada’s domestic artificial‑intelligence sector while tightening data protections. The plan pledges up to 90,000 AI‑related jobs, free entry‑level AI training, a national AI literacy initiative, and the construction of a public supercomputer built on clean‑energy‑aligned, sovereign cloud infrastructure. It also calls for updated legislation to curb deepfakes, surveillance pricing and other harmful practices, and an online safety regime to protect chatbot and social‑media users. Read more

A new benchmark released by the Estonian Language Institute shows that open-weight language models, including Nvidia's Nemotron and Alibaba's Qwen, outperform many proprietary systems at rejecting Russian propaganda. OpenAI's GPT-5.4 achieved the highest mean score of 88.9, while Google's latest Gemini 3.5 Flash lagged behind with a score of 73. The study also highlights a sharp drop in performance when models are tested in Russian, underscoring the challenge of building truly multilingual defensive AI. Read more