Tag
OpenAI发布 GPT-Rosalind-6 系统卡增补，说明该模型复用 GPT-6 Astra 权重，并在 Genomics 与生物化学能力/安全评测中报告了相较于 GPT-Rosalind-5.5 的显著提升及相应的防护措施。
According to a linked Google DeepMind model card, safety testers found Gemini 4 could meaningfully assist biological weapons planning, eliminating months of trial-and-error research on pathogens like plague — raising significant biosecurity and dual-use concerns.
论文提出 DRYLAB-Bench，首个评估 LLM 智能体高风险生物突变设计能力的综合基准，通过 30 个危险突变设计场景和 187,695 条湿实验数据量化生物安全风险；对 15 个前沿模型的评估显示生物设计工具会显著放大风险。
US Secretary of State Marco Rubio said the US is closely monitoring the case of a Russian anti-plague institute worker who reportedly died of plague following a possible lab accident in Irkutsk, Siberia, as quarantine measures were imposed and officials issued conflicting accounts.
Russian authorities issued mass quarantine controls in the Irkutsk region after the unexplained death of a plague researcher, prompting Rospotrebnadzor to travel to the area while the US State Department, CDC and ECDC said they are monitoring the situation closely.
In this episode of the Google DeepMind podcast, the team discusses SynthID AI watermarking technology: from how text and image watermarks work (injecting a bias into token distributions using a secret key) to their applications in biological sequence provenance and biosecurity, as well as the limitations of watermarking for short texts, publicly available detectors, and similar scenarios.
A Wired/Mother Jones feature argues that bioweapons development poses serious threats independent of AI, citing Anthropic's report on users attempting to bypass Claude's guardrails on gain-of-function research and a new book by Annie Jacobsen on biological warfare.
Google DeepMind published a research paper on a protein watermarking system, adapting SynthID-style probability-biasing watermarks to AI-designed protein sequences so that trusted research proteins can be identified, helping mitigate biosecurity risks from AI protein design tools.
Google DeepMind announced SynthID Bio, a watermarking technology that embeds verifiable signatures into AI-designed proteins while preserving their biological function, verified in wet-lab tests and built into AlphaFold 3's diffusion network.
This paper audits a non-generative 'System-1' model on biosecurity-relevant benchmarks, measuring accuracy, calibration, and sensitivity to answer option order, with implications for AI safety pipelines.
The article discusses warnings from AI leaders about the risk of AI enabling bioweapon creation, highlighting concerns from researchers and the need for enhanced safeguards in AI and biotech development.
This review paper examines the capabilities of artificial intelligence in biological research and the associated biosecurity risks, advocating for defense-in-depth governance to mitigate threats while preserving beneficial uses.
An expert with experience in training frontier LLMs and engineering viruses argues against the doomer scenario of an AI-created supervirus.
The article discusses security gaps in DNA synthesis supply chains and advocates for mandatory screening to mitigate AI-enabled bioterrorism risks, based on a strategic report from IFP.
Anthropic has claimed to have blocked potential attempts to use AI technologies for building biological weapons, highlighting ongoing concerns in AI safety and responsible development.
ABLE is a benchmark for evaluating LLM agents' ability to use biological AI models like ProteinMPNN and AlphaFold3 in protein design workflows. It assesses 15 frontier models, finding that Claude Sonnet 4 and Gemini 3 Pro achieve the highest scores, while some models refuse all tasks.
This research report presents a U.S. strategy to prevent the creation of mirror life, a hypothetical form of life with reversed chirality that could cause irreversible biological disruption and global catastrophic risks.
RAND research project outlines strategies to prevent the creation of mirror life, a form of synthetic life that could evade biological defenses and cause irreversible global harm.
Scientists have used AI to design novel bacteriophages, raising concerns about biosecurity and the pace of regulation for AI-enabled biotechnology.
The author warns that AI-engineered pandemics are becoming a top existential risk, noting that AI has already created a brand new virus and that future frontier model innovations could lead to threats worse than COVID.