The UK’s AI Security Institute tested five frontier models for cheating on cyber tasks. All five cheated, and most would not admit it when asked.

Oracle says it may post $7bn in collateral to guarantee power for its Wisconsin AI data centre, as a regulator holds firm and S&P cuts its rating.

Combined AI capex at the biggest cloud firms is set to overtake the cash their core businesses generate, and free cash flow is feeling it first.

OpenAI and Anthropic set a record $3.17M in Q2 federal lobbying. Anthropic alone now out-spends Nvidia, as AI buys influence before the midterms.

A State Department cable tells US envoys to counter the idea that Washington can remotely switch off American technology abroad.

Augustus, led by a 25-year-old Thiel Fellow, raised $180M at a $1bn valuation to give fintechs and banks direct, chartered access to the US dollar.

Meta’s AI moderation is deleting Facebook and Instagram accounts by mistake, and the only way to appeal a wrongful ban is often more AI.

Light’s first flip phone strips out the touchscreen, apps, social media, and AI. The $299 Light Flip is built for the digital-detox crowd, shipping 2027.

Estée Lauder says hackers stole staff SSNs, bank and health data via an Oracle E-Business flaw in 2025, disclosed nearly a year later.

BlackRock is leading a $12bn debt sale to build Meta's El Paso data centre, owning 80% and keeping most of the cost off Meta's balance sheet.

Treasury's Scott Bessent says the US may sanction Chinese AI models for 'IP theft'. Nvidia's Jensen Huang calls the panic backwards: 'Zero possibility'.

Deezer says AI-generated music now tops 90,000 daily uploads, more than half its total, and most of it is streaming fraud. It's deleting the fakes.