Related stories
Loading related stories
The Verge
Spotify says it removed more than 50 million tracks that it judged AI-generated slop or spam from recommendations over the past year, and changed its system to highlight music by human artists. It sorts candidates using listener engagement, upload patterns, metadata and outside AI-detection tools, and it will not recommend music from AI personas, though it does not label every AI track.
Wired
Chatbots follow hidden instructions, often called a system prompt, that shape what they will and will not say. About 80 WhatsApp help pages exposed text written for Meta's support bot, including instructions not to mention data protection regulators or legal documents when users raised complaints. Meta says a bug displayed the text and that it has been fixed.
The Verge
AI agents work by driving a browser the way a person would, clicking links and filling in forms built for humans. The Verge reports that Yelp, Delta, JPMorgan Chase and Charles Schwab now block them, and that Cloudflare's bot rules catch agents alongside scrapers. When that happens the assistant stalls, and sites may ask more visitors to prove they are human.
Engadget
Apple reviews each app before it reaches the store, but its rules now push more policing onto developers. Apps that host user content must filter nonconsensual intimate images, including AI-made ones, offer reporting and blocking tools, and remove anything that slips through. Apps that hide nudify features until approval risk removal and losing their developer account.