PRESENTED BY

Cyber AI Chronicle

By Simon Ganiere · 21^st January 2024

Welcome back! This week newsletter is a 6 minutes read.

Evolution at speed

We are not even at the end of the first month of the year, and the threat landscape is amazingly busy, the number of AI news is through the roof…difficult to follow. Here is what I have done this week to keep me up to date:

• Last weekend I attended Daniel Miessler AUGMENTED AI course. His approach to enhancing daily life and work with AI are both practical and inspirational. The power of mixing AI, automation and API together is real and can lead to significant efficiencies. Also whilst prompt engineering looks like the best way to trend on social media, it has a huge importance. I highly recommend his newsletter as well, Unsupervised Learning.

• At the AI House Davos, I gained valuable insights, particularly from the session on “Expanding Universe of Generative Models” and “AI: The Great Equaliser.” These sessions provided a fresh perspective on the need for a new architecture if we want to continue the innovation and the real need to focus on risk management. You can watch the replay here and explore other insightful WEF videos on AI here.

• In the cybersecurity realm, the industry is grappling with the aftermath of the Ivanti zero-day exploit. Notably, the threat actor's shift from nation-state to mass exploitation post-disclosure is concerning and demonstrate that anything internet facing needs to be patched very quickly. Additionally, Microsoft's recent encounters with nation-state actors underscore the escalating cyber threats and the fact that you can be a heavy weight in the sector but still getting compromised.

Why It Matters?

There are a few things I’m going to take with me out of the above:

• The Need for New AI Architectures: Despite rapid advancements, there's a pressing need for innovative architectures to push AI to the next level.

• Ethical Considerations and Safety: Current AI tools, while impactful, necessitate careful consideration of safety and ethics.

• Evaluating AI: The challenge in AI evaluation is significant. The Turing Test, for instance, needs evolution. Can a single test truly determine if an AI has reached human-level intelligence?

• Balancing Innovation and Immediate Action: While innovation is crucial, there's an urgent need to focus on short-term actions for evaluating and ensuring the safety of AI models.

In the news

The new ChatGPT app store may expose users to risks from malicious bots and legitimate ones that could transfer data to insecure locations. Despite OpenAI's privacy efforts, third-party GPTs in the store could misuse user data, raising concerns about security and data privacy in this expanding ecosystem.

This resource is a practical compilation of offensive ML attack techniques, focusing on actionable code rather than academic research. It's designed to aid red teams in attacking ML environments, offering tested, effective tools. The content is continuously updated, varying from polished pages to basic placeholders.

The OWASP AI Exchange is an open-source collaborative document aimed at advancing global AI security standards and regulations. It integrates various initiatives like the EU AI Act and OWASP ML Top 10, providing a comprehensive overview of AI threats and controls. Its mission is to align and foster collaboration among standardization efforts, serving as a key resource for AI security expertise and contributing to security standardization, such as the EU AI Act.

Closing Thoughts

As we navigate through the rapidly evolving landscape of AI and cybersecurity, staying informed and proactive is more crucial than ever. The insights from Daniel Miessler's AUGMENTED AI course, the thought-provoking sessions at AI House Davos, and the latest developments in AI security standards highlight the dynamic nature of our field.

Your engagement and continuous learning are vital in shaping the future of AI and cybersecurity. To stay ahead of the curve, consider subscribing to our newsletter. You'll receive regular updates on the latest trends, breakthroughs, and challenges in AI, along with expert analyses and actionable insights.

Thanks for reading!