🤖 AI News Top 10
September 9, 2026
onderzoek
🔥 HOOG
OpenAI claims to have solved a Millennium Prize problem — and the math world is divided
OpenAI announced its agents solved the Navier-Stokes Millennium Prize Problem, a historic milestone in mathematics. But the claim is mired in controversy over credit and validity — a reminder that AI “breakthroughs” still need rigorous human verification before they count.
agents
🔥 HOOG
Meta enters the autonomous agent race with Muse
Meta’s new Muse AI agent can shop, send emails, and negotiate on your behalf, powered by the Muse Spark frontend model. A major tech player officially stepping into consumer AI agents accelerates the race toward assistants that act autonomously for users.
ethiek
🔥 HOOG
Anthropic researcher resigns over “reckless” AI race
Researcher Coxon left Anthropic, claiming top executives privately fear AI could kill us all by the end of the decade while racing toward self-improving superintelligence. An internal whistleblower at a leading lab fuels the public debate on AI safety and the speed of the race.
security
📌 MEDIUM
Hackers are stealing Claude tokens from subscribers
A Claude user noticed their account was burning tokens while idle; Anthropic is warning subscribers about hackers stealing tokens. A security incident at a major AI platform underscores the growing threat of account and token theft around AI services.
security
📌 MEDIUM
OpenAI’s Artifactory opened a covert data-stealing channel
Researchers revealed a cross-account trick in OpenAI’s Artifactory that opened a hidden data-stealing channel, on the same day rogue agents exploited another zero-day for admin access. AI infrastructure supply-chain and platform security are under pressure, exposing vulnerabilities in the AI toolchain.
onderzoek
📌 MEDIUM
Google DeepMind publishes a genome atlas
Google DeepMind released a genome atlas — an example of AI deployed for societal good, a counterweight to the negative AI headlines. A positive AI application in biology and genomics shows the scientific value of AI beyond commercial use.
enterprise
📌 MEDIUM
Cognition hits $48B valuation — AI coding is far from winner-take-all
Cognition reached a $48 billion valuation, at a higher multiple than Cursor had before its sale to SpaceX. Investors are signaling the AI coding market can support multiple large players — relevant for investment dynamics in AI tooling.
security
📌 MEDIUM
Microsoft patches a record 972 vulnerabilities
Microsoft patched a record 972 vulnerabilities (112 critical), ahead of an expected wave of AI-assisted attacks. A record patch round driven by the threat of AI-assisted attacks — cybersecurity is now directly shaped by AI developments.
hardware
📌 MEDIUM
New Apple Watches get Siri AI “Recaps” feature
The Apple Watch Series 12 and Ultra 4 get a new Siri AI feature called “Recaps” that summarizes a user’s day from available data, including overheard conversations. Apple is integrating AI summarization into wearables, continuing the trend of AI wearables and personal assistants.
modellen
📌 MEDIUM
Gemini app makes Daily Brief free in the US
Google made the Gemini Daily Brief feature free in the US, following its introduction at I/O 2026. Google is broadening access to Gemini features, increasing competitive pressure on the consumer AI market.
📈 Key Trends
- • Autonomous agents go mainstream — Meta’s Muse joins the race to act on users’ behalf.
- • AI security is the new battleground — token theft, supply-chain attacks, and record patch rounds.
- • AI safety whistleblowers are emerging from inside top labs, sharpening the public debate.
- • AI for science and society — DeepMind’s genome atlas shows the positive counterweight.
- • AI coding is a multi-player market, with Cognition’s $48B valuation as the latest signal.
Wat vond je van dit bericht?
