What's new in AI governance, safety, jobs and society

Well, let me be clear. Hacking of a government website, hacking of a private sector website — if your product does that, you are responsible.

Require fresh human approval before an irreversible or high-impact action, then log the identity, request, policy decision, approval, and resulting change in a system the agent cannot alter.

Turns out, we were being unnecessarily pessimistic. Even though the planet is on fire, extreme temperatures will not be the ones to knock us out of the evolutionary game. (So much for our soggy paper straws.) Instead, we’ve decided to build something...

Which is why US lawmakers put forward legislation, dubbed the ‘Kill Switch Act’, which would require companies to have a way to shut down problematic AI tools. Tragically, the proposal was quickly shot down.

The aggregate outcome of AI's impact on global employment is likely net positive over the long term.

Third, embedding addresses a real information problem. Frontier AI risk assessment depends on non-public facts such as training details, internal monitoring data, and incident records. An evaluator that can see only what the developer chooses to send...

Just because you’ve developed a new drug doesn’t mean you can put it on the market. You have to persuade the FDA. That takes about a billion dollars’ worth of work. That is at least the requirement that should be demanded of AI.

How computational agent theory leads to a solution of the technical alignment problem, something that requires a major rethinking of how we design AI systems.

As generative AI technologies continue to evolve, clear disclosure standards to preserve transparency are essential to civic engagement. Unfortunately, Congress has failed to keep up the pace with this rapidly advancing technology, leaving voters vul...

We have slowed down reinforcement learning and training runs because we do believe that safety should be of utmost priority. Of course, we want to hit AGI as soon as possible and we want to get there first. But we also want to make sure that AGI bene...

The ICRC has called for legally binding rules that would: Prohibit those autonomous weapons that pose the greatest threat, namely: “Unpredictable” autonomous weapons, whose effects cannot be sufficiently understood, predicted and explained; and auton...

States have a responsibility – to their own populations, and to those in other states who will face the consequences of the proliferation and use of AI-enabled weapons and by conflict escalation – to set clear limits. A company cannot determine what ...

We need to be really careful with open-source on safety. I can't talk for open models. But by definition, there's more risk with them.

We’re here today to tell Flock to get out of our communities. They’re in our neighborhoods right now, and the federal government is sending millions of dollars to states, cities, and police departments to fund systems of mass surveillance, like Flock...

Embedded evaluators are independent external experts placed within frontier AI companies to provide ongoing assessments and validation of a company’s safety, security, and operational systems, practices, and claims. [...] Contracts should contain spe...