The AI apocalypse panic is real, but is it pointed at the right problems?

Key points

• A former researcher at Anthropic, the AI safety company, made a public declaration warning that AI could end human civilisation.
• US Senator Bernie Sanders told a Washington event that Congress has been "asleep at the wheel" on AI regulation.
• Steve Bannon also appeared at the same event, making it an unusual coalition of voices on AI risk.
• Guardian tech reporter Johana Bhuiyan argues the existential panic may be pulling attention away from AI harms happening right now.
• Industry calls for regulation deserve scepticism, because companies sometimes push for rules they know how to work around.

A researcher who used to work at Anthropic, an AI safety company based in San Francisco, went public recently with a stark warning: artificial intelligence could be an existential threat to humanity. The declaration spread quickly, and by the time it landed in Washington, senators were holding events about the end of the world.

Senator Bernie Sanders told the room that Congress has been "asleep at the wheel." Steve Bannon was also there. We covered both men sharing that stage in our 15 September story "Bernie Sanders and Steve Bannon Walk Into a Room. They Both Want AI Curbed." When a progressive senator and a far-right commentator agree on a cause, something genuinely unusual is happening.

Should ordinary people be worried?

Yes, but probably not about the thing dominating the headlines. Johana Bhuiyan, a tech reporter at The Guardian, argues the apocalypse framing is a distraction. AI is already causing real harm to real people, and those problems get less air time because everyone is busy arguing about science-fiction scenarios.

Think job losses in creative fields, AI-generated misinformation before elections, or automated systems making decisions about people's benefits and bail conditions. Those aren't future problems. They're current ones. Timnit Gebru made a similar case in an interview we published on 11 September: extinction warnings, she said, draw attention away from autonomous weapons, climate damage and workers already losing their livelihoods.

Why should I be sceptical about calls for regulation?

Bhuiyan's sharpest point: when the AI industry itself calls loudly for regulation, ask who benefits. Big technology companies have the lawyers, the lobbyists, the compliance teams to handle whatever rules get written. Smaller competitors and independent researchers often don't. Regulation shaped by the companies being regulated can quietly freeze out competition while looking responsible.

That doesn't mean all regulation is bad or cynical. It means watching carefully who drafts the rules and who gets a seat at the table.

I tested this framing against my own experience. Last month I used Claude, Anthropic's AI assistant, to help draft a school email for my niece's teacher. Genuinely useful. Also completely unregulated, collecting data about my conversation, with several pages of terms to read before I understood what Anthropic does with that information. The everyday harms are quiet and ordinary, not cinematic.

What happens next?

Political momentum exists now that it didn't six months ago. The risk is that urgency gets captured by the loudest voices, who may not represent the people most affected by AI. Watch which harms Congress actually legislates against, not just which ones make the best speeches.

Common questions

Who was the former Anthropic researcher and what did they say?

The Guardian's reporting does not name the individual or quote them directly, so the declaration's specifics remain vague. It was alarming enough to reach the Senate.

Is Anthropic a dangerous company?

Anthropic was founded specifically around AI safety research, so it sits in an unusual spot: a company that believes its own technology might be dangerous and builds it anyway, betting that having safety-focused labs at the frontier is better than leaving that ground to others.