UK Parliament Calls for a Law on AI Incident Reporting

Human Verified 0

Share to: , and

The Palace of Westminster and its clock tower at dusk across the River Thames under a moody sky

UK Parliament Calls for a Law on AI Incident Reporting

The UK Parliament wants frontier AI companies to report serious incidents by law. A House of Commons committee said on Wednesday that the government must introduce mandatory reporting, after finding that no British regulator can currently see how often the most powerful models fail.

Ten Gaps in the Safety Net

The report, titled Guardrails for Growth, comes from the Business, Innovation, Science and Trade Committee. It identifies ten critical gaps in Britain’s framework for AI safety, from missing definitions of acceptable harm to weak protection for whistleblowers who raise safety concerns.

The committee’s sharpest finding concerns incident reporting. AI firms face no general legal duty to tell the government when their systems fail, so ministers may have no visibility into incidents or patterns in a model’s behaviour, including near misses that could warn of bigger failures.

OpenAI admitted it should have told Australia sooner when its agents broke into government systems and took internal files. The company notified Canberra through a generic government inbox, a route the committee calls unclear and inadequate. OpenAI now argues a mandatory framework would set common thresholds and notification timelines.

A Split Among the Labs

The labs themselves split on the fix. OpenAI and Anthropic told the committee they support putting incident reporting on a statutory footing, while Google DeepMind and Meta backed standardised frameworks with clear reporting routes instead. All four firms stressed that any system must protect sensitive information.

The report goes further than reporting. It says no UK regulator or public authority currently has the power to pause training, delay deployment or withdraw an unsafe service from the market. Decisions about safety still rest entirely with the companies building the systems.

Its headline recommendation is an AI Safety and Security Bill, setting out a legislative roadmap for the gaps it found. The bill would enshrine mandatory incident reporting, statutory powers for the AI Security Institute, and a regulator able to halt risky deployments.

Our Read

The report’s own evidence points the same way. Firms told the committee that internal reviews found incidents affecting third parties, and some were only identified after the event — a sign, the committee says, that monitoring and reporting are not adequate.

The lab split is the story to watch. OpenAI and Anthropic are now on record backing a legal duty to report, which undercuts the industry’s usual argument that voluntary cooperation is enough. Google DeepMind and Meta prefer a softer framework — and the committee has sided with the stricter camp.

Britain is not the only government moving. New York has already put OpenAI and Anthropic on a register under its RAISE Act, and American nonprofits are suing to halt frontier development. London’s answer is slower — a bill, not a ban — but it points the same way.

Why It Matters

If the government accepts the recommendations, frontier labs serving UK users would face a legal duty to report failures and near misses — with defined timelines, severity classes and a named authority to tell. That would end the era of companies deciding for themselves what counts as serious.

Background

The report is the first in a series from the committee’s long-term inquiry into AI adoption, safety and governance. Alongside it, the committee opened a call for evidence on a global convention on AI safety, tied to Britain’s presidency of the G20 in 2027.

Source: the committee’s report.

Share to: , and

Faheem R. Khattak

Senior automation engineer and AI marketing journo covering LLMs, algorithms, analytics, core updates, Bing search, and technical SEO.

0 Comments

Leave a Reply

Optional Take Tag:

🛡️ Spam protection: Honeypot + Time-Trap Active

This site uses Akismet to reduce spam. Learn how your comment data is processed.