Keep ChatGPT away from kids and the mentally unwell
SAFETY
-
Building Interactive AI Characters Responsibly
By
–
Learn more: https://
runwayml.com/news/building-
interactive-ai-characters-responsibly
… -
Frontier AI Failures and Human Dependence on Advanced Systems
By
–
The planet losing IQ points when frontier AI stutters is both hilarious and accurate. We're more dependent on these systems than most people realize, and the failure modes are only going to get weirder.
-
Smart Metrics Shape Behavior in the Enterprise
By
–
Smart metrics don't just measure performance, they shape behavior. If we're deploying #AI across the enterprise, we need guardrails against metric gaming from day one. Governance isn't a brake on innovation, it's what makes transformation durable. Especially as agentic systems
-
Agents in Production: Unpredictability Beyond Traditional Software Testing
By
–
New Conceptual Guide: You don’t know what your agent will do until it’s in production With traditional software, you ship with reasonable confidence. Test coverage handles most paths. Monitoring catches errors, latency, and query issues. When something breaks, you read the
-
Musk-OpenAI Trial: Mission vs. Profit Debate
By
–
In their paper today, the authors assert that LLMs are "homogenizing" human thought and expression https://
cell.com/trends/cogniti
ve-sciences/fulltext/S1364-6613(26)00003-3
… -
Runtime interception for prompt injection attack detection systems
By
–
Where is the runtime interception piece? I'm very skeptical of any system that claims to be able to detect prompt injection attacks from looking at prompt strings because there is effectively an infinite variety of ways they might be formatted
-
Global preparedness mission seeks support and collaboration
By
–
Great news @jackclarkSF – the world is woefully unprepared. How can we support this mission?
-
PromptSonar Limitations: Non-English Jailbreaks Detection Gaps
By
–
The README says that it doesn't handle "Non-English jailbreaks" which I think means it's of very little value https://
github.com/meghal86/promp
tsonar?tab=readme-ov-file#known-limitations–v010
… Also it appears to mainly work using regular expression? That doesn't look robust to me at all
