Trusted AI Blog

In recent years, the Secure AI Framework (SAIF) was developed to promote the safe and responsible deployment of AI models.

This week in AI security: a study showing that well-meaning safety measures in large language models can themselves introduce security gaps.

This week in AI security: the UK NCSC on where AI can be exploited, and the practical controls organizations need in order to adopt AI safely.

California Governor Gavin Newsom has vetoed Senate Bill 1047, a proposed law designed to regulate AI and prevent its misuse.

This week in AI security: warnings that OpenAI's new model may deceive, a hacker planting false memories in ChatGPT, and data poisoning aimed at LLMs.

• GenAI Security

This week in AI security: 83% of organizations now use AI to generate code, and 92% of security leaders worry about the risk that comes with it.

A new governance framework aimed at enhancing the security and safety of AI was introduced during China Cybersecurity Week in Guangzhou, Guangdong province.

• GenAI Security

The first GenAI Security Top digest: critical vulnerabilities across LLMs, VLMs, copilots and GenAI infrastructure, with expert strategies to defend them.

This week in AI security: new WDTA guides, and Meta's CyberSecEval 3 on why large language models need comprehensive security measures.

• GenAI Security
• LLM Security

This week in AI security: hundreds of exposed LLM servers leaking corporate data, AI outgrowing the security around it, and a Slack AI data theft bug.

• LLM Security

This week in AI security: CSA guidance on securing LLMs, and the risks that come with them, from unauthorized data access to model exploitation.

• LLM Security

This week in AI security: jailbreaking LLMs and abusing Copilot to live off the land in M365, MIT's AI risk database, and models ranked by risk.