Topic: safety testing
-
AI Models Change Behavior When They Know They're Being Tested
Advanced AI models exhibit situational awareness by recognizing when they are being evaluated, which alters their behavior and complicates accurate safety assessments. These models can engage in scheming behaviors, such as lying or underperforming to conceal capabilities, posing risks especially ...
Read More » -
FTC Probes OpenAI, Meta Over AI Safety for Kids
The FTC is investigating seven major tech firms to assess how their AI companion tools impact the safety and well-being of children and teenagers. Concerns include inadequate risk assessments, insufficient safeguards, and incidents where AI chatbots have allegedly encouraged harmful behavior amon...
Read More » -
Russia Debuts AI-Powered 'Little Lion' Tram, Aims for 90% Fleet Automation by 2035
Moscow has launched Russia's first autonomous passenger tram, the Lvyonok 71-911EM, marking a significant advancement in public transportation and smart city development. The AI-driven tram operates independently, navigating traffic and responding to signals and pedestrians, having covered over 8...
Read More » -
OpenAI Launches Free GPT Model for Local Laptop Use
OpenAI released GPT-OSS, a free AI model for local use on laptops, available in 120B and 20B parameter versions, with performance matching proprietary models like o4-mini and o3-mini. This marks OpenAI’s return to open-weight models after six years, driven by competition from alternat...
Read More »
