OpenAI AI Updates: November 7, 2026
Fully automated. Found, summarized and published by an automated AI pipeline. How posts are labeled
1. OpenAI and Ligo Release BioBasicsBench, Where GPT-6 Astra Scores 66.7% on 100 Raw Wet-Lab Data Tasks
OpenAI introduced BioBasicsBench on November 6, a 100-task benchmark built with enzyme-design company Ligo that tests whether AI agents can interpret the raw, messy evidence of routine lab work rather than figures curated for papers. Scientists wrote the tasks from their own experiments across protein methods, imaging and cytometry, quantification and assays, nucleic acid methods and molecular cloning, with inputs such as gel and blot images, colony plates, microscopy stacks, sequencing reads, plate-reader exports, bench photos and protocol videos; rubrics give credit for numbers, reasoning and for recognizing when the evidence cannot support a conclusion, and each task passed reviews by a second Ligo scientist and an external PhD-level wet-lab scientist. Across 37 model and reasoning-effort settings with five attempts per task (18,500 results), GPT-6 Astra at xhigh effort reached a mean reward of 66.7% (95% CI 60.3 to 73.1%) against 43.8% for GPT-5.6 Sol, while using 4,704 output tokens per task versus 9,099. OpenAI’s worked failures show the gap that remains: Astra picked the wrong colony-PCR lanes (which would lead to purifying the wrong plasmid), let a correct observation about focus-dependent contrast be overridden by a generalization, and overcounted colonies by about 12%. OpenAI says tracking these capabilities will also feed its biosafety and biosecurity assessments. Source Source
2. OpenAI Partners With Hacktron to Probe Its Systems and Gives the Firm Frontier Model Access Through Daybreak Red
OpenAI announced on November 6 that it is partnering with security firm Hacktron to find vulnerabilities in OpenAI’s systems, test how those systems connect with third-party services and strengthen its defenses using frontier AI. According to OpenAI’s post as reported by The Economic Times, the arrangement also expands Hacktron’s access to OpenAI’s frontier models through Daybreak Red, the program OpenAI uses to give vetted security teams its cyber-capable models, including the GPT-6 Cyber model launched on November 5. The announcement does not give a contract value, duration or the number of findings so far. Source Source