A popular virtual therapy platform is demanding facial scans from providers and patients to continue treatment. This biometric requirement forces vulnerable people to choose between privacy and mental health care.
Character.AI users revolt against ads, usage limits, and restrictive guardrails. The platform exemplifies how AI services degrade after gaining users.
Municipalities can't figure out how to cancel surveillance contracts, so they're literally bagging their cameras. A surreal symbol of buyer's remorse in the surveillance state.
Anthropic's Series H announcement buried the real news: their revenue crossed $47 billion run-rate this month. The AI arms race is generating massive cash flows faster than anyone predicted.
Anthropic released Claude Opus 4.8 with refreshing honesty about incremental progress. The company admits there's "still more to be done" while working on models that provide similar capabilities at lower cost.
Miles hired Claude as his "personal CFO" and calls it the best financial decision in years, operating "as a McKinsey-level analyst." The tweet signals a shift from AI as assistant to AI as professional service replacement—financial analysis being particularly ripe for automation.
AI agents are becoming "every company's biggest security test," highlighting the tension between capability and control. As agents gain autonomy, the attack surface expands exponentially—not just from external threats but from agent misuse and privilege escalation.
Ethan Mollick shares Claude's self-aware comment after writing an economics paper: rating its own identity confusion at 7/10 when pretending to be an economist. This meta-cognition about roleplay suggests models are developing sophisticated theory of mind about their own performance.
DAIR.AI challenges conventional wisdom that stronger models always need "lighter harnesses," arguing that higher-capability models might actually benefit from more structure, not less. This inverts the typical assumption about AI safety scaling.