At a world-first summit in Britain on Thursday, Western allies agreed on a safety framework for new cutting-edge models, while UN Secretary-General Antonio Guterres urged for a "united, sustained, global response" to artificial intelligence risks.
The UN chief said the world was "playing catch-up" and needed to "get ahead of the wave", as he addressed the inaugural AI Safety Summit attended by various political, technology and other figures from around the world.
The two-day gathering at Bletchley Park, north of London, ended Thursday with Western governments and companies involved in so-called next-generation "frontier" AI agreeing a new safety testing regime.
Governments from G7 and EU countries, as well as Australia, Korea, Singapore, struck a deal with AI-leading companies such as OpenAI, Anthropic, Google DeepMind and Microsoft, to test their latest models before and after release.
It comes amid growing concerns around the emerging tech, ranging from job losses and cyberattacks to humans' ability to remain in control of future systems.
Guterres told the summit that AI had "possible long-term negative consequences" on everything from jobs to culture, while its concentration in a few countries and companies "could increase geopolitical tensions".
Warning it might "exacerbate the enormous inequalities that already plague our world", he demanded a "united, sustained, global strategy, based on multilateralism and the participation of all stakeholders".
Hailing the new "landmark" agreement around testing, British Prime Minister Rishi Sunak said he believed it would help "tip the balance in favour of humanity".
Sunak, who struck an alarmist tone in a recent speech warning of AI's potential dangers, also announced that renowned AI academic Yoshua Bengio would lead a team producing an inaugural report into AI safety.
"The late Stephen Hawking once said, AI is likely to be the best or worst thing to happen to humanity," Sunak told a news conference concluding the summit.
"If we can sustain the collaboration that we have fostered over these last two days, I profoundly believe that we can make it the best."
The gathering kicked off Wednesday with an agreement signed by 28 countries and the European Union acknowledging the "need for international action".
Thursday saw senior representatives from leading Western nations formally convene, with Sunak, US Vice President Kamala Harris and EU chief Ursula von Leyen in attendance.
Leading tech figures were also present — including SpaceX and Tesla CEO Elon Musk, who turned up on both days — alongside those from academia and civil society.
China, which attended Wednesday, was not invited to the more sensitive discussions on the largely behind-closed-doors second-day sessions.
The gathering agreed to further AI safety summits in South Korea and France over the coming year.
The release of ChatGPT and other generative AI systems, which are capable of quickly producing text, images and audio from simple commands in everyday language, has captivated the public and offered a glimpse into the technology's potential.
The summit coincided with the release of a "new" Beatles song produced with AI's assistance more than four decades after it was originally recorded as a demo.
London and Washington this week both announced the establishment of institutes that will carry out that work and identify and mitigate other risks posed by AI.
Harris told delegates of US efforts to help distinguish authentic government-produced digital content from AI-generated content, and to prevent the use of discriminatory AI algorithms, her office said.
UK Foreign Secretary James Cleverly told AFP earlier that countries were responding to the urgent need for a coordinated response.
"We have to move at a pace that matches the pace of technology change, we don't have a choice," he said.
"And actually what we're seeing is a willing(ness) globally, at both government level and within the commercial world, to move very, very quickly."
Musk described the summit as "timely", acknowledging AI is "one of the existential risks that we face".
"It is potentially the most pressing one if you look at the timescale and rate of advancement," he said.
G7 powers agreed Monday on a non-binding "code of conduct" for companies developing the most advanced AI systems, while US President Joe Biden also issued an executive order regulating them domestically.
In Rome, ministers from Italy, Germany and France called for an "innovation-friendly approach" to regulating AI in Europe, as they urged more investment to challenge the United States and China.
The "inside-out and upside down" Aquellum skyscraper, to be built under water, will be accessed via floating marina
Low turnout in first three phases have raised concerns about voter disinterest in election without any notable central...
Apple is reportedly bringing "biggest" software update to its new iOS 18
Check out these new train routes that will connect Europe's big cities at budget friendly rates
Police in Georgia report killing of 35-year-old woman by aggressive dogs
Barron has been under eagle eyes of Melania Trump
