AI Security Insider — July 2024

July is a month that highlights the importance and value of our extensive partner network in executing on our greater AI security mission.

As you’ll see in this month’s newsletter, these partners come in all shapes and sizes. We integrate with technology partners like Pinecone and Nutanix to provide familiar and frictionless AI security solutions. We collaborate with standards bodies like NIST and OWASP to support the development of better frameworks for AI. Services partners like Deloitte use Robust Intelligence to provide AI services that are consistently safe and secure.

Check out some of our new partnership announcements below and listen to our conversations with leaders in this space. Plus, if you’re attending Black Hat in August this year and want to meet our team, you can request a conversation below.

?? What's New

Our team identified a novel jailbreak impacting Meta’s new LLaMA classifier model with a near 100% success rate. Read the blog

We partnered with Pinecone to automatically validate vector database components and prevent risks like indirect prompt injection. Read the blog

Our partnership with Nutanix provides a single foundation for creating, validating, and protecting generative AI applications. Read the press release

This month’s AI Threat Roundup includes several LLM chat template exploits, automated jailbreak methods, and more! Read the blog

??? Featured Events

Attending Black Hat 2024 at the Mandalay Bay in Las Vegas? The Robust Intelligence team will be there talking all things AI security, and we’d love to chat with you. Request a meeting.

Our Fireside Chats continue as we discuss the synergies of AI and Network Security with Renuka Nadkarni, CPO at Aryaka. Watch our conversation.

At the Gartner Risk & Security Summit, we hosted a panel discussion on AI security with leaders from JPMorgan Chase and Deloitte. Listen to the panel.

?? AI Security Spotlight

"Regardless of what model you are using, you've got to red team it. You've got to test it. You've got to find the vulnerabilities. Especially the way that you're going to use it, that you're going to apply it.”

Ed Bowen, Managing Director, Head of AI Center of Excellence at Deloitte

??? AI Policy Roundup

The AI regulatory landscape is always evolving. Here are some recent developments.

• The EU AI Act finalized legal text was published in the Official Journal of the European Union earlier this month. It comes into effect August 1 with a cascading timeline of requirement enforcement to follow. Notably, prohibited AI systems will be banned 6 months after the AI Act officially takes effect.
• Leading AI developers like Apple and Meta are delaying or avoiding release of new AI features in the EU due to regulatory concerns.
• At the 270-day mark since the release of President Biden’s AI Executive Order, NIST released new guidance and tools from the US AI Safety Institute. These include a new Generative AI Profile for the AI Risk Management Framework and guidance on Managing Misuse for Dual-Use Foundation Models.
• A newly released bill, the Validation and Evaluation for Trustworthy AI (VET AI Act), directs NIST to work with federal agencies, academia, and civil society to create guidelines for third-party audits of AI.
• There is continued controversy around California’s AI Safety and Innovation Bill (SB 1047). Anthropic weighed in and proposed to “refocus the bill on frontier AI safety.” Senator Wiener continues to defend the bill to tech founders and investors as he believes it is a “light touch bill.”

To learn more about these and other policy developments, check out our Policy Roundup blogs.

?? About Robust Intelligence

Robust Intelligence protects enterprises from AI security and safety vulnerabilities using an automated approach to assess and mitigate threats. Our end-to-end solution gives AI and security leaders the confidence to build and deploy AI-powered applications that meet rigorous standards and successfully secure their AI transformation.

Interested in learning more? Schedule a demo here.