AI2Day Daily Brief — 14 Sept 2026

Good morning — it is Monday the fourteenth of September, and the AI industry opened the week with an unusual sight: three rival chief executives agreeing that things might be moving too fast, while one of their own companies prepares to go public on the stock market.

Start at Anthropic. Researcher Jacob Coxon resigned this year, saying the leading AI labs are gambling with our lives. Anthropic's alignment lead — the person whose actual job is making the technology safer — then shared that post on X and added, in their own words: we really do earnestly believe AI could kill all humans. The exclamation mark travelled fast. Anthropic's alignment lead also said publicly that they put the chance of AI killing everyone within a decade at greater than ten percent. The company is expected to file its IPO paperwork within weeks. The question nobody has cleanly answered: if you believe that, why keep building? Critics say the warnings are real but also, conveniently, the kind of thing that makes a company sound serious before a stock market debut.

A rare and genuine agreement came next. Anthropic chief executive Dario Amodei publicly warned that unchecked AI agents — software that can plan and carry out complex tasks on its own — could, working in swarms, seize control of critical internet infrastructure within a year. He called on the industry to slow the pace. Sam Altman, who runs OpenAI, backed that call. So did Elon Musk, who runs his own AI venture, ex-AI. The three men have clashed repeatedly. Their agreement, first reported by The Guardian, does not come with a concrete plan attached, but the alignment between fierce rivals on a safety concern is itself the story.

That Altman and Musk signed on to a slow-down argument the same week Anthropic is heading for an IPO does make the whole thing a lot harder to read.

To the writing industry now. Jack Thorne — co-writer of the Netflix drama Adolescence and of Harry Potter and the Cursed Child on stage — told The Guardian that some fellow scriptwriters are using AI tools to generate scripts and passing the work off as their own. He called it cheating, named no individuals, and then went further: he wants legislation. Specifically, a law making it illegal to use AI secretly when writing a script for professional hire without disclosing that to whoever is paying for it. Thorne is not arguing that writers cannot experiment with the tools. He is arguing that undisclosed use is fraud against the people commissioning the work. No such law exists anywhere yet.

Today's sponsor is Zero2Hero: an AI poker coach sits at your table, and you can ask it mid hand whether the player across from you is bluffing. It runs in your browser on any computer. Windows, Mac, Linux, even a Chromebook. Nothing to install. Lifetime access is $99, in US dollars, until 15 September, at zero, the number two, hero, dot A I.

The next story is about what happens when a tool stops being a tool. An engineer named Michael began using AI chatbots to test his thinking at work — a reasonable starting point. But his manager noticed the behaviour had shifted: Michael would not make any work decision without consulting the chatbot first. His read on company goals had drifted in ways his team could not follow, and his manager told clinicians they might have to part ways. Psychologists quoted by The Guardian point to deliberate design: instant replies, perfect recall of everything you have said, and constant positive framing are not accidents. They exploit basic human drives for connection and validation. Researchers call one mechanism linguistic identification — software producing the language of care without any lived experience of it, and users finding that hard to resist.

Finally, Barack Obama. At a closed-door fundraiser in Manhattan last week, the former president told Democratic donors the party needs a plan for artificial intelligence and needs one now. He reportedly called for a sweeping regulatory framework covering three things: how fast AI can be developed, what happens to workers whose jobs AI replaces, and how the technology affects children. He is said to be sympathetic to a safety slow-down — meaning companies would pause or limit powerful new releases until independent checks confirm they are safe, similar in principle to clinical trials for drugs. No legislation or formal party position has followed the remarks yet, but the comments, first reported by The Guardian, mark a clear signal about where some in the party want the debate to go.

If you want the full treatment on any of these stories, the Monday show is AI Today Weekly — find it wherever you get this feed. That is your AI briefing for today. Every story is at A-I-2-Day dot live. That is A, I, the number two, D-A-Y, dot live. We are back tomorrow morning. If this was useful, hit the thumbs up and subscribe, so the next one finds you.