On 12 September 2023 Neil Lawrence gave oral evidence to the House of Lords Communications and Digital Committee in the first public session of its inquiry into large language models. He appeared alongside Ian Hogarth (Frontier AI Taskforce), Jean Innes (Alan Turing Institute) and Ben Brooks (Stability AI). The session recording and transcript are on Parliament’s site; ai@cam summarised the appearance as ‘A moment of steerage’ for AI technologies.

Neil argued that technical capability, business models and regulation can still shape how AI develops — “we’ve got a real moment of steerage here” — but that the problems are complex and simple answers will not work. He cautioned against concentrating attention on the Frontier AI Taskforce’s £100 million over five years when UKRI’s annual budget is of order £7 billion, and when earlier public investments in AI had drawn little comparable notice. The challenge, he said, is institutional: how public bodies deploy these tools at scale, not how a small frontier programme is spun.

He warned of regulatory capture and of a path, already visible in some US debate, that treats big tech leadership as inevitable: if large firms alone control the stack, “we effectively have autocracy by the back door.” Software engineering, he suggested, still operates with guild-like power over governments — the modern equivalent of scribes before the printing press — and democracies need innovative ways to rebalance that asymmetry. Open ecosystems matter for the same reason: closed stacks protect incumbents from the kind of disruption open source once enabled against earlier software monopolies.

On accountability, he insisted that consequential decisions must stay with people who carry reputation, not with models that generate plausible text. “A large language model can never say” what a witness says when they sit before a committee with a personal history; the policy task is to empower human decision-makers, not to replace them.

The evidence sits with the group’s AI Council work on foundation-model policy through 2022–23, including the April 2023 briefing and the June letter warning against simplistic narratives.