Lords pass new AI chatbot rules

Good morning from the proud owner of a new bathroom.

Happening Today 🗓️

See you at 10: We're hosting an online panel and Q&A at 10am today, sponsored by RELX, on the government's AI-Copyright update with experts from techUK, the University of Cambridge, YouGov and RELX. Watch live here.

In Parliament: There's a backbench debate in the Commons on tackling online harms, led by Lib Dem MP Ian Sollom.

News In Brief 🩳

Adopting, not firing: More than half of UK firms are using AI, a leap from a third last year, according to research by the British Chambers of Commerce. Those firms are also reporting productivity increases and no impact on workforce size. But the minority of SMEs going deeper on AI adoption (one in ten) are more likely to expect headcount reductions.

Cash for Manchester: "Hundreds of millions" will be invested in the Manchester Digital Campus, making it a centre of government in the North West, the Treasury said overnight. It will be home to 8,800 civil servants on a disused brownfield site in the Ancoats area. The British Business Bank, meanwhile, will invest more than £150 million into high potential companies in Northern clusters.

Lords pass strict AI chatbot rules

The House of Lords rebelled against the government last night and voted for strict new rules on AI chatbots.

What just happened? Amendments to the Crime and Policing Bill, which create a new legal framework for AI chatbots, cleared the house by 203 votes to 148 just before 8pm (that's dinner time if you're a peer). They were put forward by cross-bench peer Beeban Kidron, but are unlikely to pass through the Commons and become law because of the government's large majority.

Not good enough: Kidron argued the government's attempt to close a legal loophole around AI chatbots had "no oversight, no time limit, no scrutiny". It gives the secretary of state wide-ranging power to amend the Online Safety Act (so-called "Henry VIII powers"), something many peers spoke out against. She challenged the government to "come back with something better" if it couldn't accept her plan.

Her amendments set new standards which go beyond the Online Safety Act. They make it:

-an offence to supply a chatbot which produces content promoting terror offences, threatens national security or public safety. Individual creators of chatbots are personally liable and can be jailed for breaches.

-an offence to provide generative AI search services which produce illegal content and content harmful to a child (if the user is a child)

Come out fighting: Speaking for the government, home affairs minister David Hanson said Kidron had good intentions but there were "serious practical and legal concerns" with her plan as it would only apply to UK-based companies and have an "extremely low threshold for criminality". He argued promoting legal violence such as a boxing match would be in scope.

Fancy a game? Hanson later put forward the government's alternative approach to regulating chatbots which passed at 1am by 83 votes to 64, meaning both versions will now to the Commons for MPs to vote on*. The amendments will then return to the Lords for another vote in a process known as "ping-pong".

Wait and see

Ministers no longer have any favoured options on AI and Copyright, after confirming yesterday that their previously preferred choice of "opt out" is no longer the chosen one, but publishers are warning of another danger.

Make an exception: The government's overall position is a "nothing in, nothing out" reset, as I reported Friday. It didn't rule out new exceptions to current Text and Data Mining restrictions, something rights-holders had been calling for. Indeed the final section of the report it published yesterday suggested a commercial research exception could be the way forward. It reads:

A focused exception for science and research could make AI-driven scientific research easier, accelerating the discovery of new medicines, or supporting advances in fields such as climate modelling.

Actually the worst: That has already angered the News Media Association whose chief executive Owen Meredith described it as "even more harmful" than the "opt out" option. So much for that reset, eh?

Keeping us busy: In her statement to Parliament, tech secretary Liz Kendall wrote: “We reject any suggestion that we must choose between our creative industries and the UK’s AI sector. Both are central to the government’s industrial strategy and vital to the UK’s future prosperity." She outlined four areas for the "next phase" of work. They are:

1) A consultation this summer on digital replicas

2) A taskforce to suggest proposals on labelling AI-generated content followed by a report this autumn

3) A review of the ways for creators to control their works online

4) A working group on independent and smaller creative organisations to explore how to support content licensing.

Wait and see: The government's overall conclusion is it's too early to legislate and they will watch court cases in the US and developments in the EU. The cynic would say they're making it a problem for a future government after any Labour leadership contest in May. The pragmatic view is they don't have the political capital to take solutions forward. This "wait and see" approach was criticised by techUK's deputy chief executive Antony Walker:

"With international competitors moving ahead, the UK cannot afford for this to remain unresolved," he said. "The UK has set its sights on leading the G7 in AI adoption, but that requires a clear and enabling framework for AI innovation - and getting AI and copyright right is central to that ambition."

Join us at 10am to discuss

Spotted Elsewhere 👀

A weekly round-up of interesting stuff I've read.

What I learned: Tom Westgarth has a great blog on his experience working in government last year on the AI Opportunities Action Plan and setting up the Sovereign AI Unit. It includes lots of advice on how to get stuff done.

Hear, hear: "The way we are narrating this (AI) moment is doing serious societal harm," argues Judith Dada on her excellent Substack, arguing for a fresh framing on AI.

One chart, many takes: The FT's data king John Burn-Murdoch has broken down that viral Anthropic chart.

Just FYI: The Morning Intelligence is free for everyone until Monday April 13. From that date free subscribers will still receive the newsletter each day, but will only be able to read a preview at the top. The first 30 days of all paid subscriptions are free if taken out before April 13. After that date only seven days will be free. You can upgrade anytime to a single or group subscription here.

Thanks for reading,

Tom

*The article has been updated to reflect the fact that both Kidron's and the government's amendments passed.