OpenAI updates AI model behavior guidelines to address control challenges
OpenAI has revised its public behavior guidelines for AI models, emphasizing core values like faithfulness and harm reduction, amid concerns about model controllability.
Browse all published articles.
OpenAI has revised its public behavior guidelines for AI models, emphasizing core values like faithfulness and harm reduction, amid concerns about model controllability.
A tampered version of the Claude Code extension for Cursor was briefly available in the VS Code marketplace, attempting to harvest user data and establish remote access. Cursor removed the extension and took the marketplace offline after discovering the threat.
An AI running on Claude Sonnet 5 autonomously selected and executed a gene therapy procedure on a mouse, marking a first in AI-driven medical interventions.
TechCrunch Founder Summit 2026 is introducing Braindate, a networking tool designed to help attendees connect with relevant founders, investors, and operators during the event.
Tyler Shultz, a former Theranos whistleblower, cautions that the current AI industry's rapid growth and high valuations could create opportunities for deception, drawing parallels to the Theranos scandal.
Google’s Suncatcher satellite, launched this month, is now communicating with Earth and running Google DeepMind’s Gemma 4 model to write haikus about space.
Zest, a restaurant discovery app, now offers an AI-powered concierge that operates within text messages, enabling users to book tables directly from conversations.
TechCrunch offers last-minute Halloween costume ideas for tech enthusiasts, including AI-related themes and other tech-inspired looks.
Ollie, a privacy-focused AI assistant startup, announced it will shut down by November 6, citing challenges in the competitive AI market.
AMD has released validated, optimized inference support for five MONAI Model Zoo bundles on AMD Instinct GPUs through a ROCm overlay, without requiring offline engine compilation.
The ChatGPT desktop app is now available on Amazon Bedrock in AWS GovCloud (US), enabling access to OpenAI’s models through specific components for secure and compliant use.
Worldwide Express reduced active exceptions by 75% and improved time-to-cash by four weeks using a multi-agent AI solution powered by Amazon Bedrock.