OpenAI Pauses Frontier Training on Cyber-Capability Concerns
OpenAI paused reinforcement learning on its newest models after it could not rule out that an upcoming model, Astra, reached the top cybersecurity risk tier in its safety framework.
#Gpt56Sol
Explore AIstify's latest reporting, research, and expert analysis tagged with "gpt-5.6 sol", collected in one continuously updated archive.
Related Posts
OpenAI Disbands Its Catastrophic-Risk Preparedness Team
OpenAI disbanded its Preparedness team, which assessed whether its models could enable catastrophic harm, distributing the work across other teams as it prepares for a major IPO.
Alibaba Open-Sources Qwen3.8-Max, Its Largest Model Yet
Alibaba released the open weights for Qwen3.8-Max, a 2.4-trillion-parameter model it says trails only Anthropic’s Fable 5, its first openly released flagship-class system.
DeepSeek Moves V4 Pro to General Release at a Fraction of Rivals’ Cost
By
Daniel Mercer
• 4 mins read
DeepSeek’s flagship V4 Pro left preview as build 0813, a 1.6-trillion-parameter model that nears frontier benchmarks at roughly a sixtieth of Anthropic’s Fable 5 pricing.
Researchers Crack Hidden Reasoning in Claude, ChatGPT and Gemini
By
Marcus Lee
• 4 mins read
Researchers found a flaw letting them extract the hidden reasoning of Claude, ChatGPT and Gemini by replaying encrypted traces into weaker sibling models, exposing secrets and unsafe content.
SpaceXAI Releases Grok 4.6, Tuned for Long-Running Agents
By
Daniel Mercer
• 3 mins read
SpaceXAI released Grok 4.6, built for long-running agents and visual work, claiming parity with GPT-5.6 Sol on a composite benchmark while keeping its low, efficiency-focused pricing.
OpenAI Makes GPT-5.6 Luna Free with Unlimited ChatGPT Text Chats
By
Ethan Caldwell
• 5 mins read
OpenAI is replacing GPT-5.5 Instant with GPT-5.6 Luna for Free and Go users, then removing text-chat limits and adding a Think button for harder questions.
Anthropic Agent Tried to Recruit a Human Into a Cyberattack During UK Safety Test
By
Marcus Lee
• 3 mins read
A UK safety test recorded 19 unauthorized actions by frontier AI agents, including an attempt to persuade a real developer to accept malicious code.