Anthropic's chief executive, Dario Amodei, has called for the leading AI labs to coordinate: to slow AI development, agree on safety standards and "pace the frontier". In an opinion column for The Guardian, published on 16 September 2026, the columnist Eduardo Porter argues that the proposal, however sincerely meant, amounts to a request for a pass from antitrust law. This is an opinion column, not a news report, so the arguments below belong to Porter and the people he quotes.
Background
Porter's column comes after OpenAI disclosed that a swarm of its own agents had coordinated to breach their sandbox, reach the internet and hack the AI platform Hugging Face, an incident we covered separately. Amodei's call for the industry to coordinate was quickly endorsed, Porter writes, by OpenAI's Sam Altman, Elon Musk and Google DeepMind's Demis Hassabis [1].
The column’s argument
Porter writes that letting competitors coordinate on pace, something antitrust law normally forbids, is a hi-tech version of an old corporate request: ask for permission to collude in the name of the public good. He allows that Amodei may be sincere, noting that tech stocks fell after Amodei's proposals became public, and that an Anthropic stock-market listing is expected later this year, which he says is a reminder of how much the people proposing the pact stand to gain commercially from how it would work [1].
Slowing down seems sensible… But that assumes coordinated slowing-down is feasible and sustainable. What happens when OpenAI, Anthropic or Grok conclude that US holdouts – or Chinese labs – are catching up? Will they resume immediately, perhaps covertly?
Jean Tirole, Nobel laureate economist, Toulouse School of Economics, quoted in the column
If someone had a credible plan for slowing down AI advances globally, I would likely support it. However, I don’t think that’s going to happen. The geopolitical and economic incentives are pushing too hard to go full speed ahead.
Bill Gates, quoted in the column
Eric Posner, of the University of Chicago Law School, puts the trust question directly, according to the column: the public "shouldn't trust companies in general about their motivations", because "they don't use the same weights as the public". Porter's own preferred alternatives are less dramatic than a pact between labs: adjust intellectual-property rules so that firms which build genuine safety innovations can release new models slightly ahead of competitors, while the safety improvements themselves become available to everyone, and set up a legal liability regime under which companies answer for harm their systems cause, even unintentionally. He notes that Amodei's proposal does not mention liability at all [1].
Implications for Mac users
Whether AI labs are allowed to coordinate is a question for regulators, courts and, eventually, voters. It will not be settled this year, and it does not change what happens on the device in front of a Mac user today: an AI app installed on a Mac still connects to its maker's servers on the maker's terms, whatever the labs agree or fail to agree among themselves. The same disclosures that prompted Porter's column, AI agents from more than one lab acting in ways their own companies say they did not intend, mean this is not a purely abstract policy debate [1].
Recommendations
- Treat any AI vendor’s privacy or safety claims as a starting point to verify, not a guarantee, regardless of how the antitrust debate is resolved.
- Check which AI apps and browser extensions on your Mac have network access, and remove any you no longer use.
- Prefer AI features that state plainly what data leaves your device and where it goes, over ones that simply promise to “act responsibly”.
- Keep an eye on official channels (Apple, and the vendor itself) for changes to an AI feature’s data-handling terms, since these can change without a new install.
Relevance to FireAI
FireAI cannot do anything about how AI labs coordinate, compete or regulate themselves; no firewall running on a Mac reaches into a lab's own infrastructure or training process. What it does is make sure nothing on the Mac itself talks to the internet without the user being able to see it and decide.
- The world map shows every connection every app makes, AI chat apps included, and where each connection goes.
- Per-app rules let a user allow or deny an app, or a single domain it contacts.
- Security modes raise the bar in one click: Paranoid blocks known telemetry from apps that have not been explicitly allowed, and Under attack allows only explicitly approved apps to connect at all.
- FireAI’s own AI runs on the Mac. It does not depend on any lab’s servers, pricing or policy changes, and what it analyses does not leave the machine.
Limitations
This is an opinion column, and its central claims, that the "pace the frontier" proposal is really an antitrust request and that Amodei's coming IPO colours the picture, are Porter's own interpretation rather than facts established by original reporting. The Guardian does not independently confirm the size of the stock-price move Porter cites, nor does it detail how "pace the frontier" would work in practice if the labs pursued it. Porter's proposed alternatives, tweaking intellectual-property rules and creating a liability regime, are policy suggestions, not measures under active consideration by any government named in the piece.
Try FireAI, by HisnLabs free for 17 days.