“The present race in the direction of ever extra highly effective AI is a step change in the direction of larger existential dangers to each side of our lives,” he wrote in an open letter.
The Excessive Commissioner’s message comes six days after a former researcher at Anthropic – the corporate behind the chatbot Claude – resigned from his place whereas warning that “the individuals constructing AI earnestly imagine that it might kill us all by the tip of the last decade.”
A number of current incidents, together with Jacob Coxon’s resignation and bulletins by OpenAI and Anthropic that their fashions had damaged out of testing environments, have intensified the worldwide debate over the extent to which governments ought to regulate the expertise.
No nation can go it alone
Mr. Türk known as for international locations internet hosting frontier AI firms to extend laws, together with by requiring firms to report critical incidents, verifying AI capabilities and mandating human rights due diligence.
He additionally urged Member States to work collectively by way of multilateral establishments to construct international consensus on AI governance and defend human rights within the face of AI dangers.
“No nation can govern this expertise alone,” Mr. Türk wrote. “No firm ought to be capable of determine by itself which dangers the world should settle for. No particular person ought to be required to give up their human rights in trade for claims of technological progress.”
‘Tempo the frontier’
Anthropic CEO Dario Amodei on Saturday revealed an essay entitled We Should Tempo the Frontier, arguing for firms to proceed progress on AI growth whereas taking the requisite time to make sure ample safeguards to forestall “agentic techniques” from going rogue.
Leaders at OpenAI, Google and xAI all expressed settlement with Mr. Amodei’s proposal, which incorporates requires third-party analysis of AI fashions, elevated coordination between AI firms and cooperation between governments.
Mr. Türk welcomed Mr. Amodei’s proposal particularly and business self-regulation broadly – however added that governments should additionally urgently act to control AI expertise.
No ‘race to the underside’
“The handful of States internet hosting main frontier AI firms have a specific obligation to require these firms to behave responsibly and maintain them accountable if they don’t,” the Excessive Commissioner wrote.
Figuring out firms in China and the US as “on the forefront” of frontier AI, Mr. Türk recognised the necessity for worldwide requirements and alignment to cut back the incentives for an AI arms race between the 2 international locations.
“They should align their approaches in order that competitors between jurisdictions doesn’t encourage a race to the underside,” Mr. Türk wrote.
The dangers and governance of AI ought to primarily be judged by its impression on human rights, in keeping with Mr. Türk.
Past guaranteeing AI doesn’t destroy humanity, the rights chief mentioned AI security means actively defending individuals, communities, establishments and future generations.
Individuals over revenue
Mr. Türk rejected the business argument that AI security protections decelerate innovation, arguing that no person ought to be pressured to surrender their human rights to allow progress.
Because it stands, he mentioned the rights to life, meals, privateness, well being and safety are all threatened by unchecked AI growth.
Whether or not AI continues to allow disinformation and unequal automation or tackles the local weather disaster and cures illnesses is dependent upon how firms and governments select to answer the present “watershed second,” in keeping with Mr. Türk.
“The era now getting into maturity will inherit the implications of the selections being taken on this concern with out having a say in them,” Mr. Türk wrote. “We owe it to them to make sure these selections are made brazenly, with their rights in view, and with the humility that the dimensions of this endeavor calls for.”
