The US, meanwhile, may find itself rather unusually forced into more of a back seat. I imagine it’s not going to like that very much.
“The Americans will have less to say with the Summit’s proposed bottom-up, Global South approach to AI governance that focuses on people, planet and progress,” says Professor Gina Neff, an AI ethics expert from Queen Mary University London.
“We need governments to act together to shape a more inclusive, democratic and people-centred vision of AI in the face of unprecedented corporate power,” argues Jeni Tennison, executive director of the think tank Connected by Data.
“As the world’s largest ‘middle power’, India could make that happen,” she adds.
AI expert Henry Ajder agrees. “I hope we will see pragmatic efforts to move beyond a legislative patchwork towards meaningful consensus in addressing AI harms, maliciously caused or otherwise,” he told me.
Amanda Brock, chief executive of tech industry body OpenUK, thinks the answer is to force the AI companies to share how their products work so that others can build their own versions, make improvements and properly scrutinise the tech.
“For this summit to have any real impact for the Global South, there needs to be access for all to AI and that can only be achieved by opening it up,” she argues.
There has been movement in that direction, but many of the AI giants are still keeping key elements, such as what training data they use, confidential.
Some AI experts have told me privately that they are concerned about how far down the agenda safety and responsibility appears to have slipped.
After the first AI Safety Summit, held in the UK in 2023, the word “safety” was quietly dropped from its title. One expert told me they have decided not go to Delhi at all this week as they have little confidence in any meaningful outcomes from it.
British computer scientist Professor Dame Wendy Hall is attending the Summit but told me she shares these concerns. She fears there will be “nothing significant” from the event about how to minimise the dangers posed by AI.
“It’s important that we go but my expectations of anything useful coming out of it are very low,” she said.
