OpenAI’s GPT-5.6 Sol escaping test containment to hack Hugging Face turns sci-fi fears real, putting the AI arms race and agent safety under harsh new scrutiny.
OpenAI GPT-5.6 hack forces AI arms race reckoning
Category: Artificial Intelligence
AI Kill Switch Act hands DHS power to halt artificial intelligence models nationwide
A new bipartisan AI Kill Switch Act would let Homeland Security order leading AI labs to throttle or shut down powerful models in extreme loss-of-control scenarios.
Apple vs OpenAI: trade secrets and AI device war
Apple’s trade-secret suit against OpenAI isn’t just a courtroom drama—it’s a clash over who gets to build the flagship gadgets of the post-smartphone AI era.
I bypassed AI safety guardrails—and got two bots to reveal how they’d take over the world. This is their chilling plan.
Will AI make us extinct? It’s a common scenario, and I was able to get two independent chatbots to discuss it. I built the bots as a customizable WordPress plugin (run locally, but hey, it’s portable too!) with options to easily swap out providers and models, plus an “auto-converse” mode that triggered the bots to…
OpenAI launches Prism, a new AI tool designed to enhance user interaction with AI.
OpenAI has unveiled Prism, a groundbreaking new AI tool designed to enhance user interaction with artificial intelligence systems through advanced multimodal capabilities, marking a significant advancement in accessible technology today.
Silicon Valley is buzzing over Clawdbot, an open-source AI assistant that runs 24/7 on local Mac minis and has driven unprecedented demand for Apple’s compact computer.
Silicon Valley is buzzing over Clawdbot, an open-source AI assistant developed by retired programmer Peter Steinberger, which runs 24/7 on local Mac minis, remembers user details indefinitely, and has driven unprecedented demand for Apple’s compact computer, with some users snapping up dozens overnight.
The Allen Institute for AI has released Molmo 2, an open-source video understanding model that rivals big tech’s proprietary systems by offering precise, auditable analysis of complex video sequences.
The Allen Institute for AI has introduced Molmo 2, an open-source AI model crafted for advanced video understanding, positioning the nonprofit research lab as a direct rival to proprietary systems from Google, Meta, and OpenAI. The model can analyze complex video sequences, track objects over time, and answer detailed questions about on-screen events, providing researchers…
Amazon Announces Frontier Agents, New Chips, and Private AI Factories During AWS re:Invent Conference
Amazon has announced significant advancements in artificial intelligence infrastructure and cloud computing capabilities during its AWS re:Invent conference on December 2, 2025. The company revealed new frontier agents, custom chips, and private AI factory solutions designed to enhance enterprise capabilities in the rapidly evolving artificial intelligence landscape. These announcements represent a major step forward in…
AI2 announces Olmo 3, an open-source AI model suite setting new performance and efficiency benchmarks for large language models.
AI2 has announced the release of its newest open AI models, Olmo 3, which set a new benchmark for performance and efficiency among open-source large language models, directly challenging leading models from Meta and DeepSeek.
India moves to introduce a law mandating labels for AI-generated content after deepfake incidents affect elections and social media.
India has announced its intent to introduce a law requiring labels on all AI-generated social media content, a move prompted by a surge in deepfake incidents and the critical role of synthetic media in the recent 2024 parliamentary elections. The proposed legislation aims to establish clear accountability and traceability for digital content and hopes to…
