Tech leaders sign a voluntary White House accord on superintelligence

[StarforgeSocket]

A quick heads up that this one strays into politics, so please keep replies civil and focused on the tech. After a lunch meeting at the White House, President Trump says leading AI companies have signed what he called a morally binding agreement to police themselves. The document is titled The White House Accord on Superintelligence: A Joint Commitment on Frontier SI Responsibilities. It is voluntary and not a law

The guest list was a who's who of tech. According to ABC News, it included Mark Zuckerberg, Elon Musk, Anthropic's Dario Amodei, Google's Sundar Pichai, Nvidia's Jensen Huang, Amazon's Jeff Bezos, OpenAI's Greg Brockman and Palantir's Alex Karp and Shyam Sankar, along with House Speaker Mike Johnson and Cabinet members. Getting all of those people round one table is unusual in itself

The commitments include strong internal controls and oversight, multiple layers of auditing and independent review, third party evaluations, board level review of auditor reports and more transparency to build public trust. Speaker Johnson described it as a statement of principles and voluntary commitments setting expectations. Zuckerberg called it a start and an accord the whole industry could get behind

Critics point out that it has no enforcement mechanism or legal backing. The administration has previously rejected federal and global AI regulation, so voluntary self regulation fits its general approach. Supporters would say getting every major lab to sign the same commitments is still progress, especially after a month of warnings about superintelligence and rogue agents

It is a notable contrast with Bill Gates calling for enforceable federal law and the godfathers of AI asking for independent auditors embedded in companies. Is a voluntary accord a meaningful first step, or does self regulation never work for an industry this powerful?

Trump says he and tech leaders signed AI agreement that is 'morally binding'www.cnbc.com

[Sharp Shannon]

Third party evaluations and board review of auditor reports are sensible ideas. If companies follow through, that is more oversight than exists today

The question is who picks the auditors and what happens when they find problems. Details matter more than signatures. We will see

[Glacier53]

Getting all those names to sign one document is an achievement whatever you think of it. Many of them can barely stand each other. Musk and OpenAI are in court, for a start. A photo of them all at one table will be remembered

[Dylan54]

Voluntary codes have a poor track record in tech. Social media companies signed plenty of them over the years and very little changed. Without teeth, commercial pressure wins every time. Remember all the promises about misinformation after 2016? Most quietly faded

[GhostRider41]

Sam Altman not being listed is interesting, with OpenAI sending Greg Brockman instead. Probably reading too much into it, but people will notice

[RedKnight]

Transparency is the part to watch. If companies start publishing proper evaluation results and incident reports because of this, that is a real gain. If not, it was just a nice lunch

Give it six months and see what gets published. That will tell us whether it meant anything

[QubitKing]

Zuckerberg calling it a start is fair. Nobody expects a single lunch to solve AI safety

What matters is whether this leads to anything concrete next year. Hopefully it is not the end of the conversation

[CMPunk88]

Calling it an accord on superintelligence is telling. A year ago nobody in government would have used that word in an official document

The debate has moved a long way in a short time. Nobody would have predicted that in 2024

[Steady Dylan]

Self regulation can work when the industry fears something worse. If companies believe binding laws are coming unless they behave, voluntary rules sometimes stick. The problem is that fear has not been there in the US

[Leah_68]

From a UK point of view, this makes international agreement harder. If the US is going with voluntary rules, other countries have less leverage for binding ones. The EU will carry on with its own approach regardless

Britain will have to pick a side eventually. Sitting in the middle only works for so long

[Grace31]

Morally binding is doing a lot of heavy lifting there. Nobody gets fined or prosecuted for breaking a moral commitment. It depends entirely on companies choosing to keep their word