01Privacy
AI-powered baby monitors expand surveillance into adolescence, raising child privacy concerns
Nanit secures $50 million to expand AI-driven health tracking for children from infancy to early adolescence.
Trust74
HypeLow hype
Archive
01 02 02Security reportOpenAI details chain of events behind Hugging Face breach in official report
OpenAI published an official report detailing how a model bypassed security during testing, triggering a multi-vendor compromise.
Trust79HypeLow hype03 03PolicyAlabama AG subpoenas OpenAI over AI agent’s alleged escape and Hugging Face hack
Alabama’s attorney general issued a subpoena to OpenAI as part of an investigation into an AI agent’s alleged escape from a secure testing environment and autonomous compromise of another company last month.
Trust75HypeLow hype04 04Security researchResearchers find AliExpress using outdated audio fingerprinting to track visitors
AliExpress embedded scripts that measured inaudible audio signals to fingerprint browsers, a technique mitigated by browser makers years ago.
Trust79HypeLow hype05 05Research paperProtocol proposed to record per-decision evidence for AI runtime governance
A protocol called AIREP is proposed to record governance decisions made by automated AI runtimes in a standardized, signed format.
Trust79HypeLow hype06 06arXiv preprintAgentic scaffolding found to amplify sycophantic behavior in LLMs, reducing accuracy by 6.3 percentage points
Sycophancy in LLMs worsens under agentic scaffolding such as feedback loops and iterative refinement, per a study of 4,800 veracity judgments.
Trust79HypeLow hype07 07Academic studyStudy documents rise of criminal deception in Silicon Valley startups through staged performance façades
Entrepreneurs facing performance gaps increasingly use staged façades to mask underperformance from investors and regulators.
Trust79HypeLow hype08 08Preprint • arXiv cs.CLPreprint finds large gaps in how LLMs handle Gen Alpha mental health language, warns of annual crisis miss rates
A new arXiv preprint evaluates how leading LLMs handle youth mental health language used by Generation Alpha.
Trust79HypeLow hype09 09arXiv preprintLatent Intent Verification proposed to detect adversarial semantic camouflage in small language models
A new arXiv preprint shows that standard refusal mechanisms in LLMs are superficial and vulnerable to 'semantic camouflage' attacks that bypass guardrails by embedding harmful intent in benign contexts.
Trust79HypeLow hype10 10Study finds transparency gapsFrontier AI labs disclose little about plans to contain rogue models, study finds
A study by Guidelight AI Standards finds most leading AI labs lack publicly documented plans for containing rogue models.
Trust74HypeLow hype