The Guardrails

Regulation, safety, and governance

The Take — Chinese data isn't the risk; the last fine-tune is

The Guardrails

The Take — Chinese data isn't the risk; the last fine-tune is

Nous Research took Alibaba's Qwen3.8-27B, put about fifty million tokens of supervision through it, and cut the share of answers carrying Chinese state propaganda from 22 percent to 0.9 percent without losing measurable capability. Most coverage read that as a story about Chinese censorship. I think it is a story about where a model's politics actually live — and it is not where the people writing export rules think it is. As our brief on the release laid out, Arete-27B is supervised…

METR: Anthropic's progress doubled, its own alarm may have rung in July

The Guardrails

METR: Anthropic's progress doubled, its own alarm may have rung in July

METR spent eleven weeks inside Anthropic's research pipeline and came back with the number the company has been circling for a year: its models are improving roughly twice as fast as they were before October 2025. The auditors also say something sharper than the headline — Anthropic's own early-warning threshold for automated R&D may already have been crossed on July 15, and the public risk report that said it hadn't understated the evidence. Neither finding is a claim that a machine woke up.

EU finds TikTok's For You feed breached the DSA — 6% fine at stake

The Guardrails

EU finds TikTok's For You feed breached the DSA — 6% fine at stake

The European Commission has spent two and a half years building this case, and on Friday it named the defendant precisely: not the videos TikTok hosts, but the machine that picks the next one. The European Commission has preliminarily found TikTok in breach of the Digital Services Act over the "rabbit hole effect" of its recommender system, the Commission announced on Friday. The finding covers two duties — assessing the risk and mitigating it — and concludes TikTok failed at both.

Deep Dive — Nasdaq's $2.3 trillion selloff and the 2027 capex peak

The Guardrails

Deep Dive — Nasdaq's $2.3 trillion selloff and the 2027 capex peak

Thursday was the worst day for US tech stocks since 2020, and the cause was not a missed earnings number or a bad inflation print. It was a set of county ballots. The Nasdaq 100 fell 6%, the Nasdaq Composite 5.5%, and the index closed more than 10% below its recent peak — a correction, in the technical sense. The Philadelphia Semiconductor Index dropped 12.2%, its worst session since March 2020. Bloomberg put the one-day loss of market value at a record $2.3 trillion, according to Bloomberg.

GPT-6 Cyber: 48.5% on ExploitGym, rated Critical for cyber

The Guardrails

GPT-6 Cyber: 48.5% on ExploitGym, rated Critical for cyber

OpenAI shipped a model today that is better at breaking into software than anything it has released, and it is only selling it to vetted defenders. The pitch is that the same training that made it more capable also made it more obedient. OpenAI released GPT-6 Cyber on Thursday, a security model it now rates Critical under its own Preparedness Framework — up from High for GPT-5.6 Cyber — and it is handing it out slowly, through the invite-only Daybreak Red program.

Moonshot is shopping for 8,000 Nvidia Blackwell GPUs in Europe

The Guardrails

Moonshot is shopping for 8,000 Nvidia Blackwell GPUs in Europe

Moonshot AI has spent this year trying to rent its way onto the silicon Washington will not let it buy. The Beijing lab behind Kimi talked to neocloud operators about leasing Nvidia Blackwell capacity hosted in European data centres, according to four people familiar with the matter who spoke to CNBC — clusters of 8,000 or more GPUs, three of them said. Two described the interest as inference capacity, the compute that serves finished models to users; one also knew of interest in training.

Britain's £400 billion AI growth figure came from Google

The Guardrails

Britain's £400 billion AI growth figure came from Google

For two years British ministers said AI would add £400 billion to the economy by 2030. On 26 October the AI minister was asked what the number rested on, and answered in writing: a report commissioned by Google. The government's own central estimate is £120 billion — less than a third of the promise. The answer, in a footnote The question came from Ben Spencer, the Conservative shadow Cabinet Office minister, and it was narrow: what is the evidential basis for the statement in the AI…

Data center bans swept the ballot — Texas flipped, 13 GW at risk

The Guardrails

Data center bans swept the ballot — Texas flipped, 13 GW at risk

For two years the fight over AI infrastructure was fought in planning commissions and interconnection queues. On Tuesday it was fought at the ballot box, and the industry lost nearly every vote it was in. Voters restricted data center projects in almost every local measure put before them, and Texas — the state expected to host a quarter of the American data center capacity coming online by 2030 — elected a Democratic governor who campaigned on halting new builds. 404 Media's Jason Koebler…

Google schools: AI off by default — and Search can't be had without it

The Guardrails

Google schools: AI off by default — and Search can't be had without it

Three stories about the same thing from different angles: who controls the defaults, who pays for the capacity, and who is now lending the money. Start with the defaults, because Google just moved the most consequential one in American classrooms. Google will switch its AI products off by default for K-12 student accounts, and the one product schools cannot replace — Search — comes with its AI attached.

Get SiliconNoon in your inbox

One email a day with the AI stories that matter — what happened, the numbers, and why it matters. No spam, unsubscribe in one click.