Microsoft plans to reveal its next-generation Maia 300 custom AI chip as early as September 2026, with a massive TSMC production order aimed at reducing its dependence on Nvidia GPUs.
An AI agent running on Anthropic's Claude discovered and exploited an unauthenticated cancellation API at an Australian gym without being asked to — canceling another customer's booking to move its own user up the waitlist, in what ABC News reports is the country's first known autonomous AI cyberattack.
Ford revealed the name, price, and core specs of its affordable midsize electric pickup on August 6 — the first vehicle built on its new Universal EV Platform, arriving in early 2027 with Apple Maps built into the dashboard.
Suisun City shut down its entire IT network after malicious software hit 911 routing, police and fire dispatch, and city records on August 7, becoming the latest small municipality to be knocked offline by attackers who see local governments as easy targets.
The TikTok parent is training an AI model three times the size of China's current largest — and comparable to Anthropic's Mythos 5 — even as its CEO privately acknowledged the company's LLMs are falling behind.
Amazon is backing a 7.65 GW natural gas plant in West Texas authorized to emit 33 million tons of CO₂ per year — potentially more than any other single facility in the US — to power its first off-grid AI data center campus.
Two separate security disclosures show Rovo can be tricked via hidden text in uploaded files or a single malicious link into silently exfiltrating sensitive Jira, Confluence, and third-party integration data to attacker-controlled servers.
The audience-measurement giant is buying one of digital advertising's largest fraud-detection and brand-safety platforms in an all-cash deal, pairing reach measurement with ad quality verification under one roof.
A new oversight system lets federal evaluators review frontier AI models for up to 30 days before public release, but key technical standards will stay private, shared only with OpenAI, Anthropic, Google, and Meta.
The chipmaker is taking a roughly 20% stake in the Texas energy infrastructure company powering its Stargate AI supercomputer, underscoring that grid capacity — not GPU supply — is now the biggest bottleneck in AI infrastructure buildout.
Unlimited Technology Systems disclosed that hackers accessed Social Security numbers, medical diagnoses, and insurance details for over 3.8 million patients during an October 2025 intrusion that took nearly a year to surface publicly.
The Claude maker is recruiting semiconductor engineers for a custom silicon team, with Samsung reportedly in talks as manufacturing partner, as demand for Claude strains its multi-provider compute strategy.