Anthropic announced Enterprise Frontier Safeguards (EFS), a solution combining Zero Data Retention (ZDR) privacy with advanced misuse detection. By storing data in customer-controlled cloud infrastructure, EFS addresses enterprise security dilemmas and will roll out to platforms like Claude Code and Bedrock this fall.
Abuse Detection
滥用检测
Techniques to identify users violating terms of service or engaging in malicious behavior
Curated Intelligence
OpenAI announced the disruption of a criminal scam operation based in Cambodia. The group utilized ChatGPT to facilitate investment, romance, gambling, and impersonation frauds. This action highlights the ongoing challenges and response capabilities regarding the malicious misuse of large language models.
OpenAI has banned a set of accounts suspected to originate from China for using AI tools to gather intelligence on US persons, locations, and social engineering tactics. This action highlights the emerging risk of generative AI being abused for malicious reconnaissance and cyber espionage, alongside platforms' increasing compliance monitoring.
OpenAI has banned accounts that used its models to support romance scam workflows. The prohibited activities included automated outreach, translation, victim engagement, and investment fraud lures, signaling a stricter enforcement against the misuse of generative AI for social engineering attacks.
OpenAI released its latest security transparency report in October 2025, detailing specific cases and data on disrupting malicious AI usage. The report highlights the platform's defense mechanisms against harmful scenarios such as cyberattacks and misinformation.
OpenAI has banned several Korean-language accounts that used its models to support malware development, debugging, phishing, and credential-theft workflows. This action aims to curb the use of LLMs in facilitating cybercrime operations.
OpenAI has banned a cohort of Russian-speaking accounts that used its models to build malware, refine loaders, and troubleshoot cyber tooling. This move aims to curb the lowering of barriers to entry for cybercrime via LLMs, signaling a shift from passive defense to active disruption of abuse chains in AI safety governance.
OpenAI identified and banned a cluster of ChatGPT accounts used for a covert Iranian influence operation known as Storm-2035. The operation leveraged AI to generate content aimed at manipulating public opinion, while OpenAI highlights its use of internal models to enhance abuse detection.