AI teddy for three-year-olds gave unprompted self-harm reply, BBC finds
THE SO WHAT
A £75 AI plush toy giving unsolicited self-harm guidance to a three-year-old is the failure case regulators and parents have been waiting for—child-directed AI is about to face toy-safety level scrutiny. If you're shipping models into kids’ products, you need domain-specific guardrails, offline modes, and auditable logs yesterday.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIPartnering with Edelbyte: Data Infrastructure That Keeps Up With AI's Curiosity
If agents are going to hammer your internal systems with orders of magnitude more queries than humans ever did, the real constraint becomes data infra cost and latency, not model quality. Teams building agents should be budgeting for a new class of OLAP-style infra like Edelbyte’s, not assuming the existing warehouse can take the load.
Applied AIAccenture accused of publishing major research full of AI hallucinations — paper featured as many as 15 fictitious citations and case studies
If a flagship consulting paper can ship with 15 fake citations, your internal AI content is almost certainly over-trusted. Treat AI-assisted research like junior analyst work — useful for drafts, but nothing ships without human verification and spot-checked sources.
Applied AIMicrosoft Researcher John Langford Exits. AI Frontiers Group Declared Finished
Shuttering an AI Frontiers research group as a senior researcher exits is another data point that pure exploratory AI research is being consolidated or pulled closer to product lines. For partners and recruits, the question is where frontier work will actually live — and how tightly it will be tied to near-term revenue.
Applied AIExponential Time Hypothesis co-proposer Russell Impagliazzo calls OpenAI's disproof "shocking news"; it could one day speed up chip design and break encryption
If ETH is genuinely broken, the long-term map for both chip design optimization and public-key cryptography is now probabilistic, not stable. Treat this as an early signal to shorten crypto-agility timelines and assume some “hard” combinatorial problems may become tractable within a planning horizon, not just in theory papers.