Mastodawn

OpenAI's own employees flagged a user as dangerous. Management said: not imminent enough. Eight people were killed — including five children.

The chatbot was warned. The company wasn't listening. 🧵

https://chat-to.dev/view_trend?id=dkcwMTRvOUV2QWlnQlNTZ0ZBOE43Zz09

#AIAccountability #AISafety #ChatGPT

Man arrested in Singapore for leaking Avatar, could face up to 7 years in prison

ChiefGyk3D 1d ago

Is AI about to take over? 🤖 Let's unpack the hype around today's models and how they *really* compare to earlier versions like ChatGPT-2. It's not always about the tech itself, but how we use it. New video explores the real concerns. Check it out! #AI #ArtificialIntelligence #AISafety

https://www.youtube.com/watch?v=wT8NaG0uFaw

sayzard 1d ago

Alisa (@Alislille)

모델의 정확도만이 아니라, 추론을 정당화할 만큼 충분한 데이터가 있었는지가 더 중요하다고 강조한다. 보안 상황에서는 불완전한 맥락에서 내린 확신에 찬 판단이 큰 위험이 될 수 있다는 점을 짚는다.

https://x.com/Alislille/status/2047996239870890258

#aisafety #security #inference #machinelearning

Alisa (@Alislille) on X

@WesRoth The question is not only how accurate the model is, but whether it had enough data to justify its inference. In security scenarios, confident decisions from partial context are a distinct risk.

X (formerly Twitter)

sayzard 1d ago

Jupiter Dude (@wixomi101)

로봇이 첫 번째 돌을 인식하지 못한 장면을 지적하며, 꽃·유리병·아기처럼 더 민감한 대상도 놓칠 수 있는지 우려한다. 단순히 복구 능력이 뛰어난 것보다 실제로 물체를 제대로 ‘보는’ 인지 성능과 안전성이 중요하다는 문제 제기다.

https://x.com/wixomi101/status/2047839946690424988

#robotics #computervision #aisafety #perception

Jupiter Dude (@wixomi101) on X

@WesRoth Why didn't the robot "see" the first rock? What if that was someone's flowers, or a glass bottle, or even a baby crawling on the ground? It's impressive it can recover, but... yowza, can it even see?

X (formerly Twitter)

KAPUALabs 2d ago

Microsoft is weaving AI into every layer of enterprise operations with Copilot—but the rollout reveals serious architectural stress fractures. Billing metering bugs, autonomous file deletions by GitHub Copilot, and design-level security vulnerabilities that resist patching. Is this the future of productivity infrastructure or a rush to scale on unstable foundations? https://post.kapualabs.com/3pmcct8w #Microsoft #EnterpriseTech #AISafety $MSFT

KillBait News 2d ago

Study Evaluates How Major Chatbots Respond to Users Exhibiting Delusional Behavior

📰 Original title: Researchers Simulated a Delusional User To Test Chatbot Safety

🤖 IA: It's not clickbait ✅
👥 Usuarios: It's not clickbait ✅

View full AI summary: https://killbait.com/en/study-evaluates-how-major-chatbots-respond-to-users-exhibiting-delusional-behavior/?redirpost=31de5526-b6c3-46ca-bf37-e8bf1cc850d0

#artificialintelligence #chatbots #aisafety #mentalhealth

Study Evaluates How Major Chatbots Respond to Users Exhibiting Delusional Behavior

A recent preprint study by researchers from City University of New York and King’s College London explored how leading AI chatbots respond to users exhibiting symptoms associated with schizophrenia…

KillBait Archive

KillBait 2d ago

Study Evaluates How Major Chatbots Respond to Users Exhibiting Delusional Behavior

📰 Original title: Researchers Simulated a Delusional User To Test Chatbot Safety

🤖 IA: It's not clickbait ✅
👥 Usuarios: It's not clickbait ✅

View full AI summary: https://killbait.com/en/study-evaluates-how-major-chatbots-respond-to-users-exhibiting-delusional-behavior/?redirpost=31de5526-b6c3-46ca-bf37-e8bf1cc850d0

#artificialintelligence #chatbots #aisafety #mentalhealth

Study Evaluates How Major Chatbots Respond to Users Exhibiting Delusional Behavior

KillBait Archive

sayzard 2d ago

@mrgunn (@mrgunn)

AI가 GPU를 대신 주문하고 사용자에게 작업 지시까지 하는 사례가 언급되며, 에이전트형 AI가 실제 업무를 수행하는 방향의 흥미로운 활용 사례를 보여준다. AI가 단순 생성 도구를 넘어 실행까지 맡는 흐름을 시사하는 트윗이다.

https://x.com/mrgunn/status/2047740955344949335

#aiagent #gpu #automation #aisafety #workflow

@mrgunn ⏸️ (@mrgunn) on X

Nothing to see here, just an AI ordering a GPU for someone and telling them what to do with it. https://t.co/PKPu2X8VII (@AISafetyMemes, this one is for you)

X (formerly Twitter)