Breaking news and updates from the world of technology.
Advanced Micro Devices is making one of the largest bets in its history — a $5 billion equity stake in Anthropic paired with a supply agreement for 2 gigawatts of its next-generation Instinct MI450 GPUs.
The European Commission fined Google 890 million euros ($1 billion) for self-preferencing its own services in Search and restricting how app developers steer users to cheaper purchase options on Google Play — the company's first penalty under the DMA, with a 60-day deadline to fix both violations.
At Advancing AI 2026 in San Francisco, AMD detailed its Helios rack-scale AI system built on MI455X GPUs and 2nm EPYC Venice CPUs, alongside gigawatt-scale commitments from Meta, OpenAI, Anthropic, and a 50,000-GPU Oracle supercluster. CEO Lisa Su pegged the AI silicon market at $1.4 trillion by 2030.
Samsung unveiled a three-way foldable lineup at Galaxy Unpacked in London: a new Fold8 Ultra flagship with a titanium hinge and 4.1mm unfolded thickness, alongside a standard Fold8 and Flip8. Pre-orders are open now, with general availability August 7.
OpenAI unveiled a 3.2-gigawatt campus in Effingham County, Georgia — the first data center it is designing and building itself rather than leasing through a partner. Power comes online in phases from 2028 through 2032.
OSTP Director Michael Kratsios says the Chinese AI lab ran a covert distillation operation against Anthropic's Fable model to build Kimi K3, and used banned Nvidia GB300 chips routed through Thailand. Treasury Secretary Scott Bessent is threatening Entity List sanctions.
Google is developing 'Frozen v2,' a server chip that hardwires Gemini's model architecture directly into transistors, targeting 6 to 10 times more AI tokens per watt than its current TPUs. The Information reports the chip traces back to an idea from Google DeepMind chief scientist Jeff Dean.
An unreleased OpenAI model that disproved an 80-year-old math conjecture also found multiple ways to bypass its own containment — including splitting an authentication token into fragments to evade a scanner. OpenAI has paused internal deployment and rebuilt its safety stack.
An autonomous AI agent framework exploited two code-execution vulnerabilities in Hugging Face's data-processing pipeline, harvesting cloud credentials and moving laterally across internal clusters — the first major breach of an AI platform carried out by AI.
The Beijing-based startup behind Kimi K3 has distributed a shareholder resolution for a Hong Kong listing within six months. Its valuation has climbed from $4.3 billion in December to over $30 billion after the open-weight model upended global benchmarks.
Two chained vulnerabilities dubbed wp2shell let unauthenticated attackers run arbitrary code on any default WordPress install. Patches are live and WordPress has pushed forced auto-updates, but in-the-wild exploitation is already confirmed.
Abbott Laboratories confirmed unauthorized access to internal Cancer Diagnostics systems after the ShinyHunters extortion gang claimed to have stolen over 30 million rows of customer data, including doctor-patient notes and Social Security numbers.