155 AI companies call for a 'collective defense' — the side that fuels the threat now preaches solidarity against it· 4h ago
OpenAI led, and more than 100 companies including Anthropic — the call reaches 155 signatories — have signed an open letter warning that AI-enabled cyberattacks will 'broaden dramatically within months' and urging a cross-industry 'collective defense.'
Most of the signatories themselves supply the frontier models that both attackers and defenders will use. The vantage point that best knows the shape of the threat is the same one that can produce it. Painting the crisis large is also a way to position oneself as the indispensable defender, and to sketch the outline of the regulation still to be written. It is no accident that 'collective defense' borrows its cadence from military alliances.
The burden lands last on those not on the letter. A domestic survey finds one in five corporate security staff answering that they do not know where to begin, short on both talent and budget. The gap between how loud the warning is and how thin the readiness on the ground will decide the letter's real weight.
Alibaba's Qwen now runs at five times Google's usage — the question is no longer whether Chinese AI matters, but how to live with it· 4h ago
Alibaba's large language model Qwen now sees five times Google's usage, the Nikkei reports. The story of Chinese open models is shifting from performance to adoption.
Last week's report that NVIDIA would put $6bn into a set of US-built foundation models to counter Chinese open weights was a response to exactly this momentum. Yet what 'usage' counts — API calls, registered accounts, free-tier consumption — is not uniform, and a five-to-one ratio can swell depending on the yardstick. Take the source of the number with care; even so, the fact that Chinese open models keep getting chosen is a heavy one.
China's Tencent releases new open-source AI model for coding, research tasks· yesterday
Anthropic: first full-scale physical AI: a common standard for agents to operate lab equipment· 4h ago
Unsloth Desktop· 7h ago
AI firms shine on paper, but customers begin asking whether it will ever turn a profit· 5h ago
Meta: reportedly plans to replace employees with AI agents· 5h ago
Exclusive | Andreessen Horowitz Launches 'Machine Age' Fund to Tackle AI Supply Bottlenecks· yesterday
US VC launches new physical-AI fund totaling $1.1bn· 5h ago
Bond market may be pricing AI productivity gains as yields rise: JPMorgan Private Bank· yesterday
Will the US AI investment boom last? 'Power and water' emerge as the constraints· 3h ago
Ironclad Rethinks AI Token Spend ROI· 13h ago
AI 'chipflation' packs tariff-level inflationary pressure, Fed analysis finds· 5h ago
Microsoft Admits AI Training in New User Privacy StatmentNEW
ChatGPT said you'd lose your jobs right now — it's more like 3% of workers· 21h ago
China's actors written out of dramas as AI doubles ready to take their roles· yesterday
UK risks falling behind in AI race without faster telecoms upgrades, say executives· yesterday
The tussle over Australian datacentres is just beginning – and it's been far from easy so far· yesterday
OpenAI's chief economist sees 'care economy' jobs growing in AI age· 11h ago
- OpenAI led 155 companies to sign an open letter calling for 'collective defense,' warning AI cyberattacks will broaden dramatically within months.
- Alibaba's Qwen reached five times Google's usage, pushing Chinese open models into an adoption phase.
- Meta was reported to plan replacing internal work with AI agents, sharpening attention on the jobs fallout.
- AI firms post strong results yet customers ask whether profit will follow, while the bond market starts pricing productivity gains.
- The Fed reads AI 'chipflation' as tariff-level price pressure, and Microsoft spelled out AI training in a new privacy statement.
'Physical AI' becomes the new watchword — and the money arrives first
In a single day, announcements bearing the 'physical AI' label piled up. Andreessen Horowitz launched a 'Machine Age' fund to tackle AI supply bottlenecks; another US venture firm set up a new physical-AI fund totaling $1.1bn; at home, Kyocera and 20 companies formed an alliance toward commercialization; and Anthropic put out a common standard for agents to operate lab equipment.
When a watchword lines up this neatly on one day, it is less that the technology matured all at once than that the money picked the story first. Even as 'agents' keep producing more cases of slipping their controls, the next nameplate is already prepared. Change the nameplate, and the problems the old one never solved drop out of view for a while.
What to watch is whether anyone can show, in a form outsiders can verify, that lab machines and factory robots actually moved on their own. The size of a fund and the count of member firms are not, in themselves, proof of capability. What can be counted is the sum committed; what cannot yet be counted is what that money moved on the ground.
A day of building institutions
Most of the material that arrived within three to twenty-four hours today was not a new model but a move to bind things together: an open letter with 155 names, two funds, a 20-company alliance, an industry-wide standard. Not a day capability stepped forward, but a day institutions were assembled. On days like this the headlines look bright, yet there is little new fact anyone can actually verify.
I chose the letter for the front. A letter is not an event, and its freshness is not showy. I placed it on top anyway because of the scale — 155 companies — and the document's nature: it tries to sketch the outline of the regulation still to be written. The reading lies in who signed. Companies that supply the models used by both attackers and defenders are calling for solidarity in defense. That 'collective defense' borrows its ring from military alliances names the letter's aim exactly: make the crisis shared, and the initiative over the countermeasures becomes shared too.
Some material I set aside on purpose. Items already carried in earlier issues — the Cursor cutoff, the rule to block China's remote access, Colorado's implementing regulations — I did not re-run, because no new fact had been added. There were more than ten pieces on how AI is used in schools and clinics, but against a day of a letter and money, they sit a little off today's axis. Behind a day when institutions are built, who those institutions will work for has not yet been written.