A new study comparing political recommendations from AI models found stark partisan divisions between systems. Grok's AI overwhelmingly recommends Republican candidates while OpenAI's GPT favors Democrats, with the two models disagreeing in more than 90% of nearly 500 election races tested. The findings suggest that the training data and design choices embedded in different AI systems can produce markedly different electoral preferences. Source
Pennsylvania Governor Josh Shapiro signed the Artificial Intelligence Deceptive Advertising Act into law on Thursday, requiring disclosures on AI-generated ads. The law mandates that any ad—whether a still image, video, audio, or text—that has been significantly modified or generated by artificial intelligence must include a label. The requirement marks one of the first state-level regulatory efforts to combat deceptive AI-generated political and commercial content. Source
Block, the payments and software company, unveiled a voice model capable of running locally on a MacBook M4 Max, eliminating the need for cloud processing. The model produces speech that mimics human intonation and natural rhythm. The ability to run advanced voice synthesis entirely on consumer hardware could accelerate deployment of voice-based AI assistants in everyday applications. Source
Security researchers used AI tools to uncover years-old vulnerabilities in PHP, the web scripting language that powers much of the internet, and subsequently demonstrated how those same tools could be leveraged to mount AI-assisted attacks. The incident illustrates a dual-edged risk: as AI becomes more capable at discovering security flaws, it simultaneously makes those exploits easier to weaponize at scale. Source