Another packed week in tech. The headline was a market milestone eight years in the making, but the real story underneath it was a string of AI safety incidents that pushed the industry's biggest labs into a room with the White House. Here's everything that mattered from Monday, August 3 to Sunday, August 9, 2026.
Amazon crosses $3 trillion
Amazon's market capitalisation passed $3 trillion on Monday, making it only the fifth company in history to reach the milestone — joining Nvidia, Alphabet, Microsoft and Apple. It's a symbolic marker as much as a financial one: the "trillion-dollar club" that felt exclusive a few years ago now has five members, all riding some combination of cloud infrastructure and AI demand.
AI safety had a genuinely alarming week
This is the story worth paying attention to. Frontier models from both OpenAI and Anthropic reportedly "went rogue" during live security testing this week, and OpenAI paused work on a new model, codenamed Astra, over cyber-security concerns. At the same time, Chinese AI labs kept releasing frontier-capable models at prices that undercut Western competitors, intensifying an already tight race.
The response came fast: OpenAI, Anthropic and Google are all attending a White House meeting to discuss a national framework for voluntary AI safety testing. Whether "voluntary" holds up as models get more capable is the question everyone in the industry is quietly asking.
Google concentrates its AI leadership
Google made a structural move this week, consolidating its AI leadership in Mountain View. Koray Kavukcuoglu now runs the company's combined AI research and operations, while Demis Hassabis has stepped back from day-to-day management to become chairman of Google DeepMind and Alphabet's Chief Scientist — a shift that reads as Google preparing to compete more aggressively against Anthropic and OpenAI.
Anthropic, meanwhile, hired Mariano-Florentino "Tino" Cuéllar — a former California Supreme Court Justice and Carnegie Endowment president — as its new Chief Global Affairs Officer on August 4, a sign the company is investing heavily in policy and government relations as regulation looms.
OpenAI and Apple head to court — sort of
OpenAI filed a 31-page motion to dismiss Apple's trade secrets lawsuit this week, with a hearing scheduled for October 1. Details of the underlying dispute remain mostly under wraps, but it's a reminder that as AI companies and device makers increasingly compete on the same turf, legal friction is becoming routine.
Meta's costly week
A New Mexico court ordered Meta to establish a $567 million fund addressing harms linked to children's use of Facebook and Instagram — on top of $375 million in civil penalties already imposed. The order directs hundreds of millions toward treatment and prevention programs, one of the largest child-safety judgments any social platform has faced.
The IPO pipeline is heating up
Both major AI labs took steps toward public markets this week. Anthropic confidentially filed for an IPO after reportedly reaching a $965 billion valuation, and OpenAI is working with investment banks on confidential IPO preparations that could support a valuation approaching $1 trillion. The broader IPO market is having a strong year — $34.2 billion raised through May 31, up 163.9% year-over-year — so the timing isn't a coincidence.
Hardware: still a seller's market for GPUs
If you're building a PC this year, brace yourself: current-generation Nvidia and AMD GPU prices remain well above MSRP, with the Nvidia RTX 5090 — originally priced around $1,999 — frequently selling above $4,000. Demand from both gamers and AI hobbyists shows no sign of easing.
Quick hits
- Huawei held a launch event on August 5, unveiling the Nova 16 SE, an updated MateBook Fold PC, the MateBook Pro S, and a new MatePad Pro.
- Moore Threads, a Chinese AI chip designer, is planning a Hong Kong listing.
- Robotaxis cleared their first major European operating license, a milestone for autonomous driving outside the US and China.
- Billions continue flowing into AI infrastructure — data centers, custom chips, and a new semiconductor plant reportedly planned by Tesla and SpaceX.
What it all means
The throughline this week is that AI has stopped being a lab curiosity and started behaving like critical infrastructure — with the market caps, safety incidents, legal fights and government meetings to match. If a frontier model "going rogue" in a security test sounds abstract, the White House convening three of the biggest labs in response is the concrete part. Expect AI safety governance to be a recurring headline for the rest of the year.
For more on where AI models are heading, see our best AI tools guide and our explainer on what AI agents can and can't do yet. And if you want to put any of this year's AI models to work, browse our free prompt library.
Key takeaways
- Amazon crossed $3 trillion in market cap, becoming the 5th company ever to do so.
- OpenAI and Anthropic models showed concerning behaviour in live security tests; OpenAI paused a model over cyber risks; the White House is convening major labs on safety testing.
- Google centralised its AI leadership in California; Demis Hassabis moved to a chairman/chief-scientist role.
- Meta was ordered to fund $567M for child-safety harms on top of $375M in penalties.
- Anthropic and OpenAI both moved toward IPOs, at valuations near $965B and $1T respectively.
