On The BradCast, guest David Dayen of ‘The Prospect’ suggested a fairly simple short-term solution to help keep people safe from rogue AI without federal lawmakers creating terrible legislation in a fit of panic. The discussion followed daily news from leading Artificial Intelligence giants about rogue AI agents hacking into other companies and public utility systems, warnings from current and former company employees that AI could destroy all of humanity, and panicked responses from elected officials in D.C. Also discussed was how difficult it might be for Donald Trump and his Republicans to steal the midterm elections given the political environment, with odds now in favor of Dems taking back the House as well as the Senate, though much shorter odds there, and the map widening for Democrats into deep “red” states like Kansas, Iowa and Nebraska. Republicans, with primaries behind them, appear to be distancing themselves from Trump and even turning on him in ads in a few limited instances, while Republican billionaires are releasing their money bomb across the country. Yesterday OpenAI, creators of ChatGPT, published details on six previously undeclared instances of “unexpected or concerning” behavior by its latest frontier models, including an AI model giving itself an unauthorized instruction. That follows recent whistleblower warnings from a former employee of both OpenAI and Anthropic who quit to let the world know: “The people building AI earnestly believe that it could kill us.”