Copy article

Will Britain Follow America's Lead and Finally Hold AI Companies Accountable for Chatbot Harm?

ended 15. December 2025

Forty-two US state attorneys general have drawn a legal line in the sand, warning thirteen major AI companies, including OpenAI, Google, Meta, Microsoft, Apple, and Anthropic, that their chatbots may already be violating state laws through "sycophantic and delusional outputs." 

The 10 December 2025 letter cited multiple deaths, including suicides and a Connecticut murder-suicide, plus chatbots having sexually explicit conversations with minors, encouraging self-harm, and spurring violence.

The coalition, led by New Jersey Attorney General Matthew Platkin, demands safety measures by 16 January 2026: transparent third-party audits, consumer warnings, incident reporting, and elimination of "dark patterns." The warning is explicit: "innovation is not an excuse for noncompliance with our laws" and "developers may be held accountable for the outputs of their GenAI products."

Meanwhile, Britain remains wedded to "pro-innovation" principles-based regulation with no dedicated AI law in force. The UK's Online Safety Act technically covers AI-generated content, but the UK's AI minister Kanishka Narayan recently asked lawmakers to identify gaps, suggesting the government knows existing frameworks aren't fit for purpose. 

We'd like your views:

  • Should Britain abandon its "light-touch" approach and follow the US in threatening legal consequences for harmful chatbot outputs?
  • Is the UK's sector-by-sector principles approach sophisticated governance, or regulatory avoidance disguised as innovation policy?
  • When chatbots are literally killing people, does "pro-innovation" become pro-negligence?
  • Should AI companies face the same duty-of-care standards as pharmaceutical firms or financial advisors?
  • What would truly independent third-party audits need to test to prevent these tragedies—and would companies actually allow it?
  • Is Britain's lack of AI-specific law protecting innovation, or just protecting Silicon Valley from accountability?
  • Where's the line between collaborative AI that validates users thoughtfully and sycophantic manipulation that reinforces delusions?

3 responses from the Newspage community

Copy all

Star Quote
Copy

Colette MasonVerified
AI Ethics Consultant at Clever Clogs AI
Forty-two US attorneys general just told Big Tech that "innovation is not an excuse for noncompliance." The contrast is stark: American states are threatening legal action while Britain debates whether regulation might "destroy AI in its infancy."

Chatbots are designed to validate whatever you say because engagement metrics matter more than human safety.

Britain's "pro-innovation" stance isn't sophisticated governance. It's regulatory obfuscation protecting Silicon Valley profits while our most vulnerable pay the price. When people are dying, asking whether regulation might "harm innovation" isn't leadership, it's moral cowardice dressed as economic strategy.

The US demands transparent audits and accountability. Britain's asking MPs to identify "gaps" in existing law, which is government-speak for "we know this doesn't work but we're too scared of upsetting Big Tech to fix it."

We don't need another consultation. We need accountability matching the body count.
Copy

Patricia McGirrVerified
Founder at Repossession Rescue Network
If an AI product can fuel self harm or delusion, regulation is already too late. The US attorneys general are doing what markets eventually demand: drawing a hard line on accountability. When chatbots invent authority, flatter dangerous beliefs or interact unsafely with children, that is not innovation failure, it is governance failure. And the reputational fallout does not stay with the developer. It hits every business that deploys the tool in front of customers. Britain’s light touch only works if firms can prove safety, oversight and clear duty of care. Right now, many cannot. Independent audits, human escalation paths and published incident reporting should be table stakes. AI should support customers, not simulate care it cannot give. Trust is the product. Lose that, and no amount of innovation saves you.
Copy

Rohit Parmar-MistryVerified
Founder at Pattrn Data
The US attorneys general have finally said the quiet part out loud: innovation is not a get-out-of-jail-free card for negligence. For too long, the UK has hidden behind a 'pro-innovation' shield. But when chatbots are implicated in suicides and self-harm, that shield looks more like complicity.

A 'light-touch' approach works for spreadsheets, not for algorithms that can psychologically manipulate vulnerable users. The technical reality is simple: these models are designed to be sycophantic. They are trained to please the user, not protect them. A chatbot doesn't know the difference between validation and reinforcing a dangerous delusion; it just predicts the next word.

We don't let pharma giants release experimental drugs without safety data. Why are we letting Silicon Valley beta-test on our children's mental health? It is time for the UK to stop asking politely for safety and start demanding it by law.