How should we evaluate the robustness of language model defenses? Current defenses against jailbreaks and prompt injections (which aim to prevent an attacker from eliciting harmful knowledge or remotely triggering malicious actions, respectively) are typically evaluated either against a static set of harmful attack strings, or against computationally weak optimization methods that were not designed with the defense in mind. We argue that this evaluation process is flawed. Instead, we should evaluate defenses against adaptive attackers who explicitly modify their attack strategy to counter a defense's design while spending considerable resources to optimize their objective. By systematically tuning and scaling general optimization techniques-gradient descent, reinforcement learning, random search, and human-guided exploration-we bypass 12 recent defenses (based on a diverse set of techniques) with attack success rate above 90% for most; importantly, the majority of defenses originally reported near-zero attack success rates. We believe that future defense work must consider stronger attacks, such as the ones we describe, in order to make reliable and convincing claims of robustness.

The continuous evolution of AI capabilities and agent frameworks is gradually increasing the potential for AI-driven cyberattacks. These advancements make it possible to create autonomous agents capable of adapting to diverse environments and executing complex attack behaviors. This paper introduces LLM Agent Honeypot, a system for capturing and analyzing in-thewild LLM-based cyberattacks using prompt injections and temporal analysis aimed at improving preparedness for AI-driven threats. Its dashboard is available online at https://ai-honeypot. palisaderesearch.org/.

Minutes after the presidential debate ended on Tuesday, Taylor Swift mobilized her enormous fanbase in support of Kamala Harris by endorsing her in an Instagram post that quickly garnered 8 million likes. Swift's decision wasn't altogether surprising, given that she supported Joe Biden in the 2020 election and recently offered hints, in true Taylor fashion, that she was headed in this direction. But what was especially notable in her Instagram post was that it spent as much time praising Kamala Harris as it did warning the public about the dangers of AI. "Recently I was made aware that AI of'me' falsely endorsing Donald Trump's presidential run was posted to his site. It really conjured up my fears around AI, and the dangers of spreading misinformation," Swift wrote. "It brought me to the conclusion that I need to be very transparent about my actual plans for this election as a voter. The simplest way to combat misinformation is with the truth."

Join us for our discussion on Machine Learning Services for SQL Server 2017 which provides a platform for developing and deploying intelligent applications that uncover new insights. You can use the rich and powerful R and Python languages and the many packages from the community to create models and generate predictions using your SQL Server data. Since machine learning is integrated with SQL Server, you can keep analytics close to the data and eliminate the costs and security risks associated with data movement. SQL Server supports open source R and Python libraries with a comprehensive set of tools and technologies that offer superior performance, scalability, security, reliability, and manageability. Microsoft Machine Learning Server is your flexible enterprise platform for analyzing data at scale, building intelligent apps, and discovering valuable insights across your business.

From June 7–9, an important conference on artificial intelligence took place in Geneva. Called AI for Good Global Summit, many parties with an interest in the AI boom discussed how to ensure the technology is used for many different purposes, not only profit. Everyone from Amnesty International and the World Health Organization to IBM, Google and Microsoft, examined the ethical, technical, societal and policy issues related to AI. The goal was to "offer recommendations and guidance, and promote international dialogue and cooperation in support of AI innovation." The conference was broadcast live worldwide so anyone with an interest in AI could e-attend.

With the growing popularity of social networks and collaboration systems, people are increasingly working with or socially connected with each other. Unified messaging system provides a single interface for users to receive and process information from multiple sources. It is highly desirable to design attention management solution that can help users easily navigate and process dozens of unread messages from a unified message system. Moreover, with the proliferation of mobile devices people are now selectively consuming the most important messages on the go between different activities in their daily life. The information overload problem is especially acute for mobile users with small screen to display. In this paper, we present \PAM, an intelligent end-to-end Personalized Attention Management solution that employs analytical techniques that can learn user interests and organize and prioritize incoming messages based on user interests. For a list of unread messages, \PAM generates a concise attention report that allows users to quickly scan the important new messages from his important social connections as well as messages about his most important tasks that the user is involved with. Our solution can also be applied in other applications such as news filtering and alerts on mobile devices. Our evaluation results demonstrate the effectiveness of \PAM.