HomeIndustriesUK's AI bill focuses on ChatGPT-style models

UK's AI bill focuses on ChatGPT-style models

Stay up to this point with free updates

UK Technology Minister Peter Kyle has assured major technology firms that a long-awaited artificial intelligence bill will focus narrowly on essentially the most advanced models and is not going to change into a sprawling “Christmas tree bill” to manage the fledgling industry.

Kyle told technology leaders that the unreal intelligence bill, expected later this 12 months, will focus solely on two things: making existing voluntary agreements between firms and the federal government legally binding and turning the UK's latest artificial intelligence institute (AI Safety Institute) into an independent government agency, in keeping with people briefed on the discussions.

“It's not going to be a Christmas tree bill,” Kyle told executives from Google, Microsoft and Apple in response to concerns raised that more regulations can be added to the bill in the course of the legislative process.

Kyle and Chancellor of the Exchequer Rachel Reeves met with executives from a lot of leading technology firms and investors – including Facebook owner Meta – on Wednesday morning to debate how the brand new government can support the technology and AI sectors to spice up UK growth.

Sir Keir Starmer was expected to announce a bill on artificial intelligence within the Speech from the Throne earlier this month, but he didn’t include it among the many 40 bills mentioned.

Instead, King said the Labor government would “seek to pass the suitable laws to impose requirements on those working to develop essentially the most powerful artificial intelligence models.”

Senior officials hope the bill can be ready for first reading by the tip of the 12 months. It will focus exclusively on ChatGPT-style basic models – large AI models being developed by a handful of firms that may analyze and generate text and multimedia – senior government sources said.

Chancellor Rachel Reeves will host an AI discussion round with Technology Minister Peter Kyle on Wednesday © Lauren Hurley/No. 10 Downing St

Sir Keir Starmer's AI Bill represents a departure from the strategy of former Prime Minister Rishi Sunak, who was reluctant to push for legal intervention in the event and deployment of AI models too early, fearing that strict regulations could hamper the industry's growth.

“Our fast-growing technology sector ranks third after the US and China and we’re world leaders relating to the security of AI,” Sunak told MPs.

Late last 12 months, Sunak's government founded the AI ​​Safety Institute, which examines AI models for risks and vulnerabilities.

During the UK AI Safety Summit in November, leading firms including OpenAI, Google DeepMind, Anthropic, Amazon, Mistral, Microsoft and Meta signed a “landmark” but non-legally binding agreement with governments including those of the UK, US and Singapore.

Under the agreement, signatory governments could review firms' latest and future models for risks and vulnerabilities before they’re made available to businesses and consumers.

These firms made further voluntary commitments in Seoul earlier this 12 months. Among other things, they promised to not “develop or use any model in any respect” if serious risks couldn’t be mitigated.

Senior UK government officials say there’s an urgent must make these voluntary agreements legally binding to make sure that firms which have already signed as much as the agreements cannot avoid their obligations if business reasons require it.

A consultation on the contents of the bill is resulting from begin in the following few weeks and is predicted to last about two months, senior officials say.

Turning AISI into an independent institution would strengthen its role as an independent body and provides firms the reassurance that the federal government will not be “respiratory down their necks”, a senior official said.

Starmer's government wants AISI to take a number one role in setting global standards for AI development that may very well be utilized by governments world wide.

Additional provisions to combat and protect against potential harms related to AI, including using mental property to coach models without permission or compensation, can be considered individually from this bill, they added.

Video: AI: blessing or curse for humanity? | FT Tech

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Must Read