August Brought AI Rulemaking, Funding, and a Few Cautionary Tales
Even Canadian Politicians Seem to Be Using ChatGPT for Their Speeches
Canadian politician appeared to read an AI chatbot's instructions aloud during a June speech on whether the province should create an energy advocate position. The moment was caught on mic in front of the legislature and became a viral example of AI creeping into public life in ways its users don't always intend.
Links:
A "Kill Switch" Bill Emerges Amid Autonomous Hacks
Two U.S. House members, from California and Texas, introduced the "AI Kill Switch Act," which would require AI companies to maintain the ability to shut down, throttle or suspend their models and would authorize the Secretary of Homeland Security to order a slowdown of any AI offering that could cause "catastrophic harm."
Links:
A Federal AI "Voluntary Framework" Is Completed on Deadline
A voluntary regulatory framework for advanced AI models, called for by a June executive order, was reportedly finalized on schedule and is now being implemented, according to the White House. The framework asks companies to give the government early access and involvement, and to help determine which of their models qualify as national-security-relevant "frontier models", though there are currently no official plans to make it public.
Link:
The Government Doubles Down on AI Investment
The Trump administration announced that 15 federal agencies will receive $5 billion to tackle long-standing scientific problems using AI, from identifying the root causes of chronic diseases to accelerating drug discovery and developing longer-lasting building materials. Involving departments including Health and Human Services, Energy, Transportation, Defense and Interior, the initiative also signals a shift toward funding individual scientists and AI over traditional university research.
Links:
Connecticut Makes Headlines With Prompt Injections in a Legal Filing
A self-represented plaintiff suing the New York Bariatric Group hid instructions in tiny 3-point white text throughout a court filing, directing any AI that processed it to side with him. When 404 Media fed the filing to ChatGPT, the model flagged and ignored the injection and ruled against the motion; the judge allowed the case to proceed but barred the plaintiff from filing electronic documents going forward.
Links:
Comments