AI This Week

OpenAI is working on a smartphone — a significant reversal from previous reports — and is fast-tracking its first "AI agent phone," with mass production targeted as early as the first half of 2027, according to supply chain analyst Ming-Chi Kuo. Drivers behind the accelerated timeline reportedly include OpenAI's planned initial public offering and intensifying competition in AI agent phones. The device would use a customized MediaTek Dimensity 9600 built on TSMC's N2P node, featuring an upgraded image signal processor for enhanced HDR, a dual-NPU architecture, LPDDR6 memory, and UFS 5.0 storage. Luxshare Precision Industry is named as the exclusive manufacturing partner. Instead of apps, the smartphone could rely on AI agents to complete tasks — bypassing the restrictions Apple and Google currently impose on system access. Kuo argues that fully controlling both the operating system and hardware is the only way for OpenAI to deliver a comprehensive AI agent service, projecting combined 2027–2028 shipments of around 30 million units. OpenAI has not publicly confirmed the device.

‍

Code leaks reveal Anthropic is building a proactive AI assistant named Orbit. It reads personal work accounts while you sleep. The system scans connected apps and delivers customized daily briefings. Orbit integrates directly with Gmail, Slack, GitHub, Calendar, Drive, and Figma. It operates asynchronously across varying time zones. OpenAI released a similar product called ChatGPT Pulse last year. However, Anthropic actively targets software engineering teams by incorporating coding and design platforms directly into the daily sync. The feature currently exists as a hidden settings toggle within recent software builds. Developers found background code confirming the development. Application activity suggests the launch is approaching fast. The company kicks off a developer conference in San Francisco on May 6.

The Pentagon announced agreements with seven leading artificial intelligence companies to deploy their technology on classified military networks, pointedly excluding Anthropic amid its ongoing dispute with the Defense Department. SpaceX, OpenAI, Google, Nvidia, Reflection, Microsoft, and Amazon Web Services will be integrated into the Defense Department's most sensitive classified systems, used for mission planning, weapons targeting, and other purposes. Anthropic was designated a supply-chain risk to U.S. national security — the first such action ever taken against an American firm — after the company refused to grant the Pentagon unrestricted access to its Claude models for use in fully autonomous weapons and mass domestic surveillance. Defense Secretary Pete Hegseth declared that "effective immediately, no contractor, supplier, or partner" doing business with the Pentagon may engage in any commercial activity with the company.

General Motors is making one of the boldest AI moves in automotive history. GM is rolling Google Gemini out to roughly four million U.S. vehicles through an over-the-air update — one of the largest deployments of generative AI into consumer hardware ever attempted. Eligible vehicles are those from model year 2022 onward across Cadillac, Chevrolet, Buick, and GMC, provided they shipped with Google built-in already integrated into the infotainment system. The upgrade is a quantum leap from old-school voice commands. Google Assistant required drivers to phrase requests in a specific format. Gemini is designed for natural language — ask it something the way you would ask a person, without memorizing command structures. Gemini reads and summarizes incoming texts, drafts replies, inserts emojis on request, and translates messages into other languages. Commercial and fleet drivers gain the most targeted tools — Gemini helps work-truck drivers optimize multi-stop routes, find the cheapest fuel, and locate trailer-friendly parking.

Elon Musk took the stand for the second day in the landmark trial that pits the world's richest person against Sam Altman, a fellow OpenAI co-founder he accuses of betraying promises to keep the company as a nonprofit dedicated to humanity's benefit. The trial centers on the 2015 birth of the ChatGPT maker as a nonprofit startup primarily funded by Musk before evolving into a capitalistic venture now valued at $852 billion. Musk fired back at OpenAI's lawyer in a tense cross-examination during the second day of the long-awaited trial. Musk, who invested about $38 million in OpenAI from December 2015 through May 2017, gave his account of OpenAI's early years — recounting how by late 2022 he was convinced Altman was trying to "steal the charity." OpenAI's lead lawyer William Savitt tried to undermine Musk's credibility, highlighting Musk's ample opportunity to object to Microsoft's involvement and multi-billion dollar investments before filing his lawsuit in 2024. Savitt also pressed Musk on whether he ever proposed that OpenAI form a for-profit arm, pointing to emails and meeting notes where Musk appeared to say OpenAI needed a for-profit aspect to compete against Google. The exchanges escalated enough for Judge Yvonne Gonzalez Rogers to intervene several times. Musk will return to the stand to complete his testimony. The trial is expected to last about four weeks.

Google plans to invest up to $40 billion in Anthropic as tech giants ramp up their strategic investments in the leaders in artificial intelligence. The deal is massive. Alphabet is committing $10 billion now, at a $350 billion valuation for Anthropic, with another $30 billion to follow if Anthropic hits certain performance targets. It's not just cash — the agreement also gives Anthropic access to a much larger pool of Google Cloud infrastructure, with Google Cloud providing 5 gigawatts of computing capacity over the next five years. The deal mirrors a similar agreement Anthropic struck with Amazon just weeks earlier — $5 billion upfront and up to $20 billion more tied to commercial milestones — making clear that both cloud giants are racing to lock up access to Anthropic's technology. Meanwhile, Anthropic's annualized revenue has soared from $1 billion at the end of 2024 to $9 billion at the end of 2025, to a stunning $30 billion as of early April 2026. The company is also reportedly considering an IPO as soon as October.

China's National Development and Reform Commission (NDRC) has blocked Meta's $2 billion acquisition of Manus, an agentic AI startup founded by Chinese engineers that relocated to Singapore before Mark Zuckerberg scooped it up late last year. The move marks one of China's most significant interventions in a cross-border deal, extending well beyond U.S.-China tensions and into the broader AI industry. With no explanation offered, the NDRC ordered both parties to unwind the deal entirely, citing laws and regulations prohibiting foreign investment in the Manus project. Around 100 Manus employees had already moved into Meta's Singapore offices as of March, with CEO Xiao Hong reporting directly to Meta COO Javier Olivan. Complicating the unwind further, Manus CEO Hong and Chief Scientist Yichao Ji are reportedly under exit bans, preventing them from leaving mainland China.

OpenAI and Microsoft announced a revamped partnership agreement that will allow the AI company to cap revenue share payments and serve customers across any cloud provider. The deal reshapes one of tech's most powerful alliances. Revenue share payments from OpenAI to Microsoft will be "subject to a total cap," but will continue through 2030, "independent of OpenAI's technology progress." Crucially, OpenAI will keep paying a revenue share to Microsoft, but Microsoft will stop paying one to OpenAI. Microsoft has agreed to drop its exclusive right to sell OpenAI's AI models — a perk that helped boost the software giant's cloud sales in the early years of the AI boom. In exchange, Microsoft will no longer pay a revenue share on OpenAI products it resells. Microsoft will continue to have a license to OpenAI's intellectual property on AI models through 2032, although the license will no longer be exclusive. Microsoft also no longer needs to determine its response if OpenAI reaches artificial general intelligence (AGI). The news follows a major OpenAI-Amazon partnership formed in February, with Amazon agreeing to invest up to $50 billion, and OpenAI expanding its existing $38 billion AWS agreement by $100 billion over eight years. Microsoft shares fell roughly 2% in premarket trading on the news.

Buyers scooping up coveted Anthropic shares have vaulted the AI giant's valuation on some trading platforms to $1 trillion, eclipsing OpenAI's market capitalization by over $100 billion in a sea change in the AI race. The investor scramble has ratcheted up in recent weeks, as demand for OpenAI shares simultaneously sags. Anthropic's annualized revenue jumped from $9 billion at the end of 2025 to $30 billion by March 2026 — a 233% increase in just one quarter — driven primarily by strong demand for its coding tools. On private stock marketplace Hiive, Anthropic shares have surged 211% to roughly $900 over the past three months, while OpenAI shares have risen just 8.5% to $608 over the same period. One Silicon Valley dealmaker even proposed swapping his 14-acre estate for Anthropic shares. One Anthropic shareholder recently offered to unload shares at a $1.15 trillion valuation.

Mozilla shipped Firefox 150 with fixes for 271 security vulnerabilities surfaced by an early version of Anthropic's Claude Mythos Preview — the largest single-cycle total the browser team has recorded. It follows an earlier collaboration in which Claude Opus 4.6 led to 22 fixes in Firefox 148, a more than tenfold jump between model generations. Firefox CTO Bobby Holley described the disorientation of seeing that volume arrive at once, noting that against a hardened target a single bug of that kind would have been an emergency a year earlier. He also drew a more reassuring conclusion: nothing Mythos surfaced was beyond the reach of an elite human researcher, which suggests the model is finding what was previously missed rather than a new class of AI-exclusive flaw. Mozilla fixed 423 security bugs across its April releases in total. Mythos remains unreleased publicly, available only to a vetted group of organizations through Anthropic's Project Glasswing program.