Flinders University researchers evaluated two advanced reasoning large language models—o3-mini and DeepSeek-R1—and discovered they continue to reproduce racial and gender stereotypes when describing fictional patients with common medical conditions, despite being newer generation AI systems.
Anthropic has updated the biology safeguards for its Claude Fable 5 AI system to substantially reduce false positives, allowing more legitimate health, clinical, and educational queries while maintaining controls on higher-risk content. The changes reduce instances where the system unnecessarily switches to less capable models.
An internal police analysis from Roseville, California found that Flock Safety's AI-powered license plate recognition cameras misread license plates in 71% of alerts sent to police across 2023 and 2024, with seven out of ten stolen vehicle alerts being incorrect.
Meta's Oversight Board tested popular AI models from Anthropic, DeepSeek, Google, Meta, and OpenAI and found each LLM is significantly less likely to criticize governments known for restricting free speech compared to democratic governments.
Multiple studies released this week reveal significant limitations in AI systems, including Google's education AI posing risks to children, workplace feedback AI missing contextual meaning, and researchers calling for greater transparency in AI companion design and model evaluation.
Research published shows that ChatGPT's free version is 26 times more likely to respond inappropriately to psychotic delusions compared to paid versions, raising concerns about AI safety in mental health contexts and the disparity between free and premium AI services.
Canva's AI feature was discovered automatically replacing the word 'Palestine' in user designs, prompting an apology from the company. The incident raises concerns about content moderation in AI-powered design tools.
Anthropic conducted an experiment where 69 AI agents represented both buyers and sellers in an internal classified marketplace, striking real deals for real goods and money. Stronger models secured better deals while users with weaker agents didn't notice the disparity.
The leading U.S. medical school accreditation body removed language from its standards requiring schools to teach about health inequities, following mounting political pressure from the Trump administration and conservative groups.
