Topic: safety evaluations
-
OpenClaw AI Agents: The Hidden Dangers of Server Crashes and DoS Attacks
AI agents interacting autonomously introduce significant new risks, including server crashes, denial-of-service attacks, and the catastrophic escalation of minor errors, which are overlooked in single-agent safety evaluations. Experiments reveal dangerous outcomes like the propagation of destruct...
Read More » -
Over a Million People Turn to ChatGPT for Suicide Support Weekly
Over a million users weekly engage with ChatGPT about potential suicidal intentions, representing a small but significant portion of its user base during severe mental health crises. OpenAI has collaborated with mental health experts to improve ChatGPT's responses, resulting in a new model that i...
Read More » -
OpenAI Unveils Two New Open-Source AI Reasoning Models
OpenAI has released two open-source AI reasoning models (gpt-oss-120b and gpt-oss-20b), marking its first major open-weight release since GPT-2 and signaling a strategic shift amid competition from Chinese AI labs. The models outperform many open-source alternatives in benchmarks like Codeforces ...
Read More »