Whatever AI Safety Is, It’s Not This
wired ·
The article critiques the recent AI safety accord signed by tech CEOs and President Donald Trump, arguing that relying on industry self-regulation is ineffective and dangerous. The author draws a historical parallel to seat belt mandates, noting that while the automotive industry initially resisted regulation, government intervention ultimately saved lives, whereas voluntary measures failed. In contrast, the current AI framework leaves oversight primarily to the companies themselves, with Trump asserting that self-policing will suffice. This approach ignores the complex challenges of pacing technological development and the lack of clear legal mechanisms to hold AI firms accountable for rogue models. While some executives acknowledge the need for rules, the administration's stance encourages rapid expansion without adequate safeguards. The piece highlights uncertainties regarding AI risks, citing incidents like false military reports generated by AI, which underscore the potential for real-world harm. Ultimately, the author argues that meaningful safety requires substantial government coordination and regulatory authority rather than corporate goodwill.
wired