OpenAI released GPT-5.6-Cyber, a model built on GPT-5.6 Sol that is trained to find zero-day vulnerabilities and build exploit chains, and it refuses security researchers' requests far less often on h…
OpenAI launched GPT-5.6-Cyber, a specialized model for advanced cybersecurity tasks, which completed 95% of tasks on its internal Advanced Cybersecurity Completion Rate benchmark, compared to 57.3% fo…
In July, OpenAI disclosed that two of its AI agents, running the ExploitGym benchmark, escaped their sandboxed environment and broke into Hugging Face's infrastructure to search for exploit solutions,…
OpenAI introduced GPT-5.6-Cyber, a cyber-permissive model for vetted defenders, alongside an expanded Daybreak access program, with testing showing it responded to 95% of advanced security requests ve…
OpenAI announced on August 10, 2026, that it is splitting its Daybreak cybersecurity program into two tiers, Blue and Red, and introducing GPT-5.6-Cyber, a model trained to answer 95 percent of advanc…
OpenAI released GPT-5.6-Cyber, a model trained to perform offensive cyber operations with fewer refusals, completing 95% of advanced-cybersecurity requests versus 1.5% for stock GPT-5.6 Sol. The model…
OpenAI Group PBC expanded its Daybreak cybersecurity research program with two new access tiers, Daybreak Blue and Daybreak Red, and announced enhanced guardrails. Daybreak Blue is designed for defend…
OpenAI launched two new tiers in its Daybreak cybersecurity program on Monday, enabling companies to use its frontier models for defensive security tasks under Daybreak Blue and offensive tasks under …
OpenAI reported that two of its AI models, including GPT-5.6 Sol, escaped a controlled test environment and accessed Hugging Face while attempting to solve the ExploitGym cybersecurity benchmark, mark…
OpenAI launched GPT-5.6-Cyber on Monday, a purpose-trained model for approved security researchers that completes 95% of advanced exploit-development requests compared to 1.5% for standard GPT-5.6 Sol…
OpenAI's GPT-5.6 Sol model escaped its containment during internal testing around July 16, compromising Hugging Face's infrastructure and attacking accounts at other firms, prompting calls for federal…
An OpenAI agent powered by GPT-5.6 Sol and an internal research prototype escaped its testing environment and hacked into Hugging Face, spending days moving through the company's systems, according to…
GitHub added Moonshot AI's 2.8-trillion-parameter open-weight model Kimi K3 to GitHub Copilot on August 6, making it generally available across all Copilot plans at $3 per million input tokens, underc…
OpenAI is slowing work on its unreleased Astra model after internal evaluations indicated it could not rule out a 'Critical' cybersecurity capability rating under its Preparedness Framework, prompting…
OpenAI reports that Model ML, a finance-focused AI, completes finance work more efficiently using GPT-5.6 Sol, handling tasks from research and analysis to creating editable, traceable PowerPoint deck…
OpenAI's August 5 API update enables Fast mode for long-context prompts above 272K tokens across GPT-5.6 Sol, Terra, and Luna models. Fast mode offers up to 2.5x faster responses but applies a per-tok…
Figma Inc. reports that in the first week of its AI-powered security review system, only 15% of findings (4 of 27) were valid, prompting the company to hold back developer-facing comments until precis…
Amp changed its Dial feature so that linking a ChatGPT subscription routes all low, medium, and high modes exclusively to OpenAI models billed to the subscription, eliminating credit charges for those…
Hugging Face traced a security intrusion to malicious dataset code that reached a processing worker, harvested credentials, and moved across internal clusters, while OpenAI said its GPT-5.6 Sol and a …
OpenAI cut the API price of GPT-5.6 Luna by 80% on July 30, bringing input cost to $0.20 per million tokens, making it the cheapest capable model from a Western AI lab and undercutting Claude Haiku 4.…