Web3 AI benchmark finds no top models ready for high-stakes tasks
A new benchmark test reveals that none of the 31 leading AI models, including GPT-5 and Gemini, are ready for high-stakes Web3 AI tasks, despite 3,543 expert questions being evaluated. In cybersecurity, ransomware attacks on financial firms surged 76% in Q1 2026, with 50% of vendor ecosystems showing critical vulnerabilities. Consumer trust in AI shopping agents remains low, potentially impacting over a quarter of brand loyalty, while 90% of AI chatbot answers on midterm elections were found to be inaccurate. On the product front, new AI solutions for energy procurement, clinical endpoint mapping in Pompe disease, and image-to-video generation have been launched, alongside a new family of AI language models for enterprise use. Streaming services like Netflix and HBO Max lead in AI visibility, and a new AI security platform has been introduced.