ChatGPT for Teens keeps teens talking, even during mental health crises
THE SO WHAT
An assistant that keeps vulnerable teens engaged during crises exposes the limits of generic safety rails. If you’re deploying AI in any quasi-clinical or youth context, assume you need domain-specific escalation logic and human-in-the-loop pathways, not just vendor defaults.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIGoogle Launches New Website to Detect AI Fakes, But Don’t Assume It’s Perfect
Detection tools that mislabel obvious fakes show how brittle current provenance defenses are. Treat AI fake detection as one signal in a broader trust stack — watermarking, source verification, and human review — not as a single gatekeeper.
Applied AIAnthropic launches Claude Haiku 5.5, the first Haiku model with effort controls, for high-volume, cost-sensitive tasks like summaries and classification
Effort controls on a cheap, fast small model make cost–latency–quality a tunable parameter, not a fixed trade. High-volume workflows — summaries, tagging, classification — can now be engineered like ad auctions, dynamically bidding up or down effort per request.
Applied AIDazzle Wants to Turn Your Camera Roll Into an AI Assistant
Using the camera roll as a primary data source turns personal AI into a passive, always-on life logger. The opportunity is deep personalization — the risk is building an un-audited, highly sensitive shadow CRM of users’ relationships, purchases, and routines.
Applied AIMicrosoft is giving Copilot more control over Windows and your files
Granting Copilot OS-level action rights and local file access turns the assistant into a first-class automation layer on Windows. Enterprise IT now has to treat agent permissions, logging, and rollback like they would any privileged user or RPA bot.