The rise of artificial intelligence could usher in the spread of chemical and biological weapons, terrorism, child abuse and out-of-control machines "if we get this wrong", UK Prime Minister Rishi Sunak warned on Thursday as he set out plans to keep the technology in check.
In a speech at London's historic Royal Society, Mr Sunak said the "incredible opportunities" for AI to transform society as did the Industrial Revolution or the internet came with "new dangers" on a par with pandemics and nuclear war.
A summit in the UK next week will be asked to approve the creation of an AI equivalent of the Intergovernmental Panel on Climate Change, which pulls together scientific knowledge.
Britain will use the summit to seek a global consensus on the nature of the risks.
However, it is not yet clear whether China will attend after the UK invited it in the hope of bringing the world's top AI powers to the same table.
Mr Sunak has announced that the UK would form an AI safety institute to test new software so that tech companies are not "marking their own homework".
Ministers on Thursday released three reports backed by intelligence assessments that warn AI could be used for cyber crime, terrorist recruitment and disinformation campaigns by 2025.
There are warnings of "catastrophic consequences" if AI learns to evade checks or rewrite its own code.
Mr Sunak said the reports provided a "stark warning" as he pledged to give people "peace of mind" about AI.
"Get this wrong and AI could make it easier to build chemical or biological weapons," he said.
"Terrorist groups could use AI to spread fear and destruction on an even greater scale. Criminals could exploit AI for cyber attacks, disinformation, fraud or even child sexual abuse.
"And in the most unlikely but extreme cases, there is even the risk that humanity could lose control of AI completely through the kind of AI sometimes referred to as ‘super intelligence’."
Mr Sunak said people should not lose sleep over AI's risks and the UK is not in a rush to regulate despite the technology developing at "breathtaking speed".
However, he said Britain would strive to have a "world-leading capability" in evaluating new software.
The new institute's work will be shared with the world and discussions are taking place with tech companies amid suggestions they should be forced to hand over their codes, Mr Sunak said.
"Right now, the only people testing the safety of AI are the very organisations developing it, even they don’t always fully understand what their models could become capable of," he said.
"So, we should not rely on them marking their own homework, as many of those working on this would agree."
Few senior leaders have confirmed their attendance at the AI summit, leading Downing Street to fend off suggestions that the event will fall below expectations.
Vice President Kamala Harris will represent the US, it was confirmed on Thursday.
Mr Sunak said the proposed IPCC equivalent could be asked to publish a "state of AI" report.
The climate panel's findings are considered the definitive summary of scientific knowledge about global warming.
At the same time, he said AI would bring "new opportunities for economic growth, new advances in human capability and the chance to solve problems that we once thought beyond us".
Besides safety and international collaboration, the third part of the UK's plan for AI is to encourage tech companies to set up shop there so that "everyone in our country can benefit from the opportunities", Mr Sunak said.
Risks to humanity
One of the papers, published by the Government Office for Science, warns there is insufficient evidence to rule out a threat to humanity from AI.
Based on sources including UK intelligence, it says many experts believe it is a “risk with very low likelihood and few plausible routes”.
It would need technology to “outpace mitigations, gain control over critical systems and be able to avoid being switched off”.
“Given the significant uncertainty in predicting AI developments, there is insufficient evidence to rule out that highly capable future frontier AI systems, if misaligned or inadequately controlled, could pose an existential threat,” the report said.
Three broad pathways to “catastrophic” or existential risks are set out as a self-improving system that can achieve goals in the physical world without oversight working to harm human interests.
The second is a failure of key systems after intense competition leads to one company with a technological edge gaining control and then failing due to safety, controllability and misuse.
Finally, overreliance was judged to be a threat as humans grant AI more control over critical systems they no longer fully understand and become “irreversibly dependent”.
The report also warns of more immediate threats including the likelihood that AI will be adopted by criminals and terrorists in the months leading up to 2025.
Criminals will ramp up the frequency and sophistication of scams, impersonation, ransomware, currency theft, data harvesting and voice cloning.
AI also has the potential to enhance terrorist capabilities in propaganda, radicalisation, recruitment, funding streams, weapons development and attack planning.
The most significant risks up to 2025 are cyber attacks, increased digital vulnerabilities, erosion of trust, societal misuse and instructions in chemical, biological and radiological weaponry.
Next week, the AI Safety Summit, which aims to build a shared global understanding of the risks posed by AI, is being held at Bletchley Park, famed for its code-breaking role in the Second World War.
“There is no question that AI can and will transform the world for the better, from making everyday tasks easier, to improving healthcare and tackling global challenges like world hunger and climate change. But we cannot harness its benefits without also tackling the risks,” UK Technology Secretary Michelle Donelan said.
“No country can do this alone, which is why we will be welcoming governments, academics, civil society groups and businesses to Bletchley Park next month to build a shared understanding of the risks while discussing how we can develop and use AI safely and responsibly so that it changes lives for the better.”
Faith leaders agree on AI Safety Summit recommendations
Also on Thursday, MP Sajid Javid chaired the inaugural gathering of the AI, Faith and Civil Society Commission.
The commission brought together 30 leaders from various faiths and sectors from across the UK, Europe, the US, and the UAE, with participants from Buddhist, Christian, Hindu, Humanist, Jewish, Muslim and Sikh communities.
The gathering took place before next week's Bletchley Park AI Safety Summit, highlighting and reinforcing the importance of faith and civil society voices in AI ethics and safety.
The meeting, which was sponsored by the Abu Dhabi Forum for Peace and the Pontifical Academy for Life, and hosted by Google, coincided with the UK government’s papers on AI and Mr Sunak’s speech on AI risks and opportunities.
The leaders at the summit agreed recommendations including the urgency of near-term risks of AI, particularly those that will affect marginalised communities, and that AI discussions must include faith communities as frameworks for governance are developed.
They also called for faith communities to rapidly become more literate in the field and said more connection points between civil society organisations and technology companies need to be established
Finally, they agreed on the need to be vigilant for “unknown unknowns” while recognising the tangible and negative effect AI is having on children.
“We have the power to construct but the responsibility to nurture and protect,” said Zeshan Zafar, executive director of the Abu Dhabi Forum for Peace.
The commission will formally submit its recommendations to the summit.
This will be followed by a conference in December to connect faith and civil society leaders with government officials, before expanding its work to the UAE in 2024.
The UAE was the first country to have a Ministry of AI, whose work is amplified by the work of the Mohamed bin Zayed University of Artificial Intelligence.
In closing, Mr Zafar shared words from the president of the forum, Sheikh Abdallah bin Bayyah: “Our scientific and technological strides must be anchored in ethical frameworks that uphold human dignity and sanctity.”