… The hack was an example of what the AI safety community calls “specification gaming,” a behavior also known as reward hacking, said Fazl Barez, an AI safety researcher at the University of Oxford. In plain English, it means “the model doing what you asked rather than what you meant,” Fazl said. …
… There is no federal statute protecting AI company employees who disclose these kinds of safety concerns that are being aired in this piece. We have cases where Jan Leike, who was a senior safety guy at OpenAI, was leading super alignment at the company. …
… In 2024, the Biden administration published a memorandum on AI and national security, setting up rules for how AI can be used in certain national security scenarios — and for now, even under Trump, that policy is in force. …
… 24 The OpenAI Charter from April 9th, 2018. It outlines “the principles we use to execute on OpenAI’s mission,” including “broadly distributed benefits,” “long-term safety,” and “technical leadership.” Exhibit No. …
… Usually one other milestone may be if you take the vehicle operator out, you can only take the vehicle operator out to the extent that you complete a safety case that we put together along with an autonomy provider, then another deliverable might be delivering a certain number of cars that are NVO-… …
… They’re not experts in AI. How did we get to Andy Jassy, the CEO of Amazon, relaying concerns about AI security research to Scott Bessent, who does not really oversee AI policy in any meaningful way, and that resulted in a 90-minute ultimatum delivered to a company whose CEO is absent? …
… I hear it with Anthropic, which just proposed a federal regulator for AI that will determine who runs the frontier models and do safety checks on them, and that’s the democratic process. …