Advanced AI Models Spark Global Cybersecurity and Financial System Fears

Thumbnail Image

The information displayed in the AIM should not be reported as representing the official views of the OECD or of its member countries.

The release of advanced AI models like Anthropic's Claude Mythos and OpenAI's GPT-5.5-Cyber has triggered global concern among regulators and financial institutions. These models can autonomously identify thousands of critical software vulnerabilities, raising fears of systemic cyberattacks and financial instability. Authorities in the US, Australia, and Singapore have issued urgent warnings and called for enhanced cybersecurity measures.[AI generated]

Why's our monitor labelling this an incident or hazard?

The article explicitly mentions an AI system (Claude Mythos) with advanced autonomous capabilities that could be used maliciously to disrupt critical infrastructure, causing harm to communities and property. While no incident of actual harm is reported, the credible risk of such harm occurring is emphasized, including government warnings and preventive measures. The AI's potential for autonomous harmful actions and its use by hackers to exploit vulnerabilities fits the definition of an AI Hazard, as it plausibly could lead to an AI Incident. The article also includes complementary information about governance and mitigation efforts, but the primary focus is on the credible risk posed by the AI system, not on a realized incident or solely on responses.[AI generated]
AI principles
Robustness & digital securitySafety
Industries
Digital securityFinancial and insurance services
Affected stakeholders
BusinessGeneral public
Harm types
Economic/PropertyPublic interest
Business function:
ICT management and information security
AI system task:
Event/anomaly detection

Articles about this incident or hazard

智通财经APP获悉，OpenAI周四推出其最新人工智能模型的衍生版本GPT-5.5-Cyber，目前正以有限预览版的形式向经过审核的网络安全团队开放使用。就在一个月前，竞争对手Anthropic推出的Claude Mythos预览版吸引众多投资者及美国政府官员高度关注。OpenAI在官方博客中表示，GPT-5.5-Cyber预览版并非旨在大幅提升网络能力，而是通过定向训练，放宽了安全类任务的执行限制。该公司已于上月末发布通用版GPT-5.5。OpenAI 称，这款网络安全专属模型，可让合规资质团队更便捷地运用其最新大模型，开展漏洞识别与分级研判、补丁有效性验证、恶意软件分析等工作流程。而通用版 GPT-5.5 内置的安全防护机制会大幅增加这类专业作业的操作难度。OpenAI在博客文章中表示，GPT-5.5-Cyber面向少数合作方开放，方便其研究高级安全工作流程，适配对专属访问权限有特殊要求的专业场景。上月 Anthropic 发布 Mythos 模型之际，同步启动名为“玻璃翼计划”(Project Glasswing)的网络安全专项计划，仅向少数企业开放模型权限。Anthropic首席执行官达里奥・阿莫迪与特朗普政府高级官员会面，讨论了该模型的技术实力与潜在影响力，尽管该公司在几周前刚刚被五角大楼列入“供应链风险”黑名单?。据报道，美国国防部要求 Anthropic 允许五角大楼不受限制地将其模型用于所有合法目的，而 Anthropic 则希望确保其技术不会被用于全自动武器或大规模监控。Mythos早期测试信息一经披露便引发市场震荡。今年3月下旬，随着该模型可自主挖掘漏洞、突破现有网络防御体系的潜在风险逐步浮出水面，市场避险情绪快速升温，网络安全板块集体承压，CrowdStrike、Palo Alto Networks、Zscaler 等多只相关个股大幅下跌。Mythos也引发美国监管层高度警惕。据报道，上个月，美联储主席杰罗姆・鲍威尔和美国财政部长斯科特・贝森特与华尔街大行高管举行闭门会议，讨论Mythos可能带来的网络安全威胁。此外，美国副总统万斯和贝森特还提前与主要科技公司高管进行了电话沟通。值得注意的是，Anthropic与OpenAI竞争已然进入白热化阶段。原本深耕企业级 AI 赛道的 Anthropic，近期大举强攻消费市场，多款实用产品迅速走红;还借助超级碗广告高调造势，将 Claude 定位为 ChatGPT 无广告替代方案，直接正面叫板 OpenAI。叠加与美国五角大楼相关舆论带来的热度加持，Claude 排名一路飙升，稳居美国苹果应用商店免费榜第二位，夹在榜首 ChatGPT 与第三名谷歌 Gemini 之间，进一步对 OpenAI 形成强势冲击。

2026-05-08
证券之星