British lawmakers are calling for a new law and a statutory oversight body to address the “serious threats” posed by AI systems to people’s human rights.
Parliament’s Joint Committee on Human Rights said the UK needed stronger safeguards including a statutory AI oversight body and new legislation to address the risks of AI.
The report’s publication follows calls from US AI leaders yesterday to slow down the pace of development of AI to give companies more time to focus on safety.
In recent weeks, incidents of AI going rogue, including a swarm of 700 AI agents created by OpenAI, collaborating to hack the open source platform Hugging Face, have led to calls for greater scrutiny of AI.
The cross-party group of MPs and peers called for the government to take “immediate” action, including banning the most dangerous uses of AI and to introduce strong sanctions against companies and organisations that damage human rights through their use of AI.
Regulation not fit for purpose
Alex Sobel, chair of the joint committee, said that AI was moving with such speed and complexity that its impact is hard to predict. “Nowhere in the world, including the UK, has a current legislative and regulatory approach that is fit for purpose,” he added.
Yesterday Anthropic founder and CEO Dario Amodei posted an essay on social media, urging a slow-down in the speed of development of AI models. “I have become convinced that fully addressing the risks requires even more prudence,” he wrote, in a post supported by Elon Musk and OpenAI CEO Sam Altman.
His comments came after Anthropic researcher Jacob Coxon publicly resigned, warning that there was a global arms race underway in AI. “The people building AI earnestly believe that it could kill us all by the end of the decade,” he wrote in a post on X.
Call for AI Bill
The cross-party committee is calling for the UK to bring forward an AI Bill, first announced in the King’s Speech in 2024, to regulate the development of powerful AI models.
It said that the rights of people in the UK were threatened by a lack of transparency over how AI systems were being used in decision-making and the lack of effective mechanisms to challenge decisions made by AI that have adverse effects on people.
The MPs and peers, in a 100-page report, call for the UK to create a statutory AI oversight body that would have the powers to test and evaluate high-risk uses of AI before they are deployed, and to prevent AI systems that pose “unacceptable” risks to human rights being deployed.
It would also have powers to investigate breaches of law and mandatory codes of conduct issued by the proposed oversight body, and to sanction any companies or individuals that fail to comply.
Companies that develop powerful AI models would be required to submit them for review, evaluation and testing to the AI Security Institute (AISI), which would also be placed on a statutory footing.
Significant risks
“The Hugging Face incident illustrates the significant risks that can arise from the newest and most advanced AI models, and legislation should address those risks on a precautionary basis with a view to avoiding harm,” the committee said.
The committee found that the current legal framework is fragmented and leaves gaps which fail to address human rights risks. It said there was “insufficient transparency” over how AI systems are used in decision-making and a lack of effective mechanisms to challenge those decisions.
Organisations that deploy AI systems that could have significant impacts on individuals or groups, should have an obligation to state that AI is being used and to explain what it is being used for, the parliamentarians said.
Human in the loop not enough
“The mere presence of a ‘human in the loop’ would not be enough to show significant human involvement or intervention in AI decision-making. It should be necessary to show that the person is sufficiently informed and independent to be able to reach an objective view that has not been improperly influenced,” their report stated.
Low-risk AI systems, used to automate simple tasks, for example planning events or shopping, should be subject to fewer and less demanding legal requirements.
But high-risk practices, which could include AI systems that profile people to predict the likelihood of committing a criminal offence, infer emotions in the workplace, or use real-time biometric identification in public places, should be prohibited, subject to a public consultation.
Alex Sobel, who introduced a ten-minute rule bill last week to require the UK government to secure an international agreement to prevent the development of superintelligent AI, said individuals should be told when AI is used to make decisions about them.
“We also want to make sure that if something does go wrong then avenues of redress will be available. We need these protections now,” he added.
The government last week rejected calls for an AI Kill Switch, proposed by Lib Dem peer Tim Clement-Jones as an amendment to the Cyber Security and Resilience Bill (CSRB), currently going through Parliament.
A government spokesperson told Computer Weekly, “Any future measures will be evidence-based and focused on the risks we need to address.”
