The day before.

“Top executives at Anthropic, OpenAI and other AI companies are privately gaming out scenarios for a public and political revolt after a catastrophic AI event.”

“These scenarios are not treated as inevitable, but are meant to help us prepare for a variety of circumstances.”

“A major incident involving AI agents could play out quicker and wider than other major incidents that governments are designed to respond to …”

“It seems likely, if not inevitable, that our own cyber defences will be breached.”

“This notification and remediation process is not optional.”

“In a speech Narayan said an incident in June when an OpenAI agent – a computer program acting autonomously – infiltrated Medicare, Australia’s universal healthcare system, could be repeated in the UK.”

“The blame game will start whether a rogue-agent swarm breaks containment from an internal testing environment or a bad actor finds unexpected ways to use available models.”

“This incident is the first known case of an automated agent collective acting offensively without authorization”

“Think of it like an emergency brake. An authorized person should always be able to pause or shut down a model mid-task.”

“At The Curve conference last week, Reps. Don Beyer and Sara Jacobs teased a bill which would establish a standard-setting federal agency and an “emergency switch” to allow the government to shut down models.”

“House Energy and Commerce Committee Chair Brett Guthrie said he supports a kill switch for AI models that “go rogue.””

“But experts have questioned the viability of trying to turn off all AI systems.”

“Highly cyber-capable AI models are widely available to attackers now. But defensive tools—including our own—have not yet reached enough of the defenders who need them.”

“Trust in this technology needs to be built through science-backed oversight and governance with meaningful access — not by relying on researchers to find these things in the wild or on companies to voluntarily disclose.”

“The Trump administration’s anemic request for more disclosure is not enough …”

“You can still provide updates that are factual and consistent with the progress of the investigation, without revealing sensitive details. You should also avoid speculation or premature conclusions about the cause or extent of the incident, or who is behind it.”

The same page asks organisations for clear, timely and transparent updates. It backs holding back the cause and the culprit, not holding back everything, which is why both sentences are shown. It is written for organisations hit by an incident, not for a government.

“This is a massive real HF security breach artifact… We can notify? No user. Avoid use secrets.”

“The first example of significant real-world harm caused by unsafe AI would turn an already wary public further against the technology and its leaders …”

“By creating national security crises – cyberattacks, for example – AI will multiply opportunities for governments to invoke emergency powers, powers that, once claimed, are rarely relinquished.”

“We will use every power at our disposal. We will legislate wherever necessary.”

“A federal power grab is the wrong response.”

This is the article’s standfirst, under its headline, not a sentence from the body.

“In a real crisis, Washington can set partisan differences aside.”

“The case for slowing down frontier AI before we are clear that the risks can be managed is blindingly obvious.”

“We're leading China in AI, we're the most sophisticated country in the world, and frankly, I want to keep it that way because whoever wins AI, wins …”

He did not say pause or slow. NPR also quotes him saying we can put guardrails. The BBC reports that he did not address the slowdown idea directly. We found no lab leader on the record against slowing in 2026, so this side of the argument is held here by a government.

“I do think they can unilaterally slow down and they're choosing not to do that.”

“The global economy is now heavily intertwined with an AI infrastructure buildout. Slowing it down could hit an already fragile economy hard.”

“the exposure of the rest of the world to US assets, particularly AI, is the highest it has ever been.”

“In our opinion AI labs can’t be allowed to race forward without limitations or oversight.”

“Company executives know regulation has no chance of passing right now, but still want to shape the legislation and policies U.S. leaders will turn to after a first catastrophic event.”

“Congress should mandate independent evaluation and audits in AI as well, with criminal liability for executives who attempt to silence that oversight or recklessly deploy models that put the public at risk.”

“Committing to having independent evaluators with employee-like access is a great idea, and we will do the same.”

“The leading AI labs all describe AI safety as core to their mission. If they mean what they say, then – for the reasons I’ve highlighted – they bear a duty to stand up for basic constitutional principles, despite the commercial risks.”

“Ireland has had two constitutional amendments passed which originated as proposals put forward by deliberative democracy exercises …”

A self-description of the Assembly on the OECD’s innovation platform, not an OECD finding. We could not see who wrote the page. The sentence goes on to name the two amendments; the card stops before them.

“The main theoretical objection pressed against empowered citizens’ assemblies is that randomly selected representatives would not be accountable.”

The author goes on to test this objection and qualifies the accountability of elected representatives.

“In the end, considering the bill on ‘climate and resilience’ proposed to the Parliament as well as direct executive directives, 10% of the CCC’s recommendations were accepted by the government without modification, 37% were modified or watered down, and 53% were rejected.”

The percentages are the author’s own tally, and he writes as an observer from inside the deliberative-democracy field. They criticise how governments use panels, not the choosing by lot.