MPs renew calls to bring generative AI into UK Online Safety Act

TL;DR:

• The House of Commons Science, Innovation and Technology Select Committee — in its response to the government’s children’s wellbeing consultation — has urged ministers to revisit its July 2025 report on misinformation and algorithmic harm, and to bring legislation forward in the new parliamentary session.
• The Committee maintains that the Online Safety Act is “already out of date” because it fails to address generative AI, and reiterates its five principles: public safety, free and safe expression, platform responsibility, user control and algorithm transparency.
• The government has so far rejected the Committee’s recommendation to extend the Online Safety Act to generative-AI platforms, arguing AI-generated content is already covered — despite Ofcom telling MPs the legal position is “not entirely clear” and that “more can be done”.

This is the second major piece of UK parliamentary pressure on generative AI in 12 months and the Lewis Silkin analysis lays out the political stakes cleanly. The Committee’s underlying argument is that the Southport-attacks-driven misinformation crisis of 2024 exposed structural gaps in the Online Safety Act regime — and that those gaps now widen with every capability jump in generative-AI image, video and voice synthesis. Government’s “AI is already in scope” position relies on a reading of the Act that even Ofcom is not willing to defend in committee evidence.

Five principles, two unresolved disputes

The Committee’s five framing principles — public safety, free expression, platform responsibility, user control, transparency — are designed to be politically durable; the harder ground is in the specifics. Two recommendations remain unresolved between government and committee: bringing generative-AI platforms inside the Online Safety Act on the same basis as other high-risk online services; and reforming the digital-advertising business model that monetises algorithmic amplification of harmful content. On the first, the government’s stated position is that no fresh legislation is needed; on the second, ministers say the matter is “under review”. For UK regulators (Ofcom, ICO, Competition and Markets Authority) the underlying technical problem is the one the AI Safety Institute described this week — current oversight techniques rest on contingent properties of today’s models that future architectures may erode (see [our AISI coverage]).

The Commons Education Select Committee’s parallel response sharpens the tone further. It is calling for a statutory ban on social-media use by under-18s and a restriction on “addictive” design elements — infinite scrolling, disappearing messages and algorithm-driven feeds — for that age group. That is much closer to the Australian regulatory direction than the UK Online Safety Act in its current form, and a meaningful test of whether Labour will tilt UK platform regulation further toward enforcement.

Looking forward

If the government does bring forward online-safety legislation in the new session, expect the scope-of-AI question to be the central drafting debate. For UK platforms and AI vendors, the practical signal is that “we are already in scope” cannot be the long-term answer — fact-checker-driven content-deprioritisation duties, transparency disclosures and algorithmic audit obligations are all live policy options. For UK SMEs in trust-and-safety, misinformation detection and AI content provenance (think C2PA-aligned products), the parliamentary pressure is procurement-relevant: platforms anticipating tighter duties will need supplier ecosystems that map to those duties.