This study investigates how individuals' perceptions of artificial intelligence (AI) limitations influence organizational readiness for AI adoption. Through semi-structured interviews with seven AI implementation experts, analyzed using the Gioia methodology, the research reveals that organizational readiness emerges through dynamic interactions between individual sensemaking, social learning, and formal integration processes. The findings demonstrate that hands-on experience with AI limitations leads to more realistic expectations and increased trust, mainly when supported by peer networks and champion systems. Organizations that successfully translate these individual and collective insights into formal governance structures achieve more sustainable AI adoption. The study advances theory by showing how organizational readiness for AI adoption evolves through continuous cycles of individual understanding, social learning, and organizational adaptation. These insights suggest that organizations should approach AI adoption not as a one-time implementation but as an ongoing strategic learning process that balances innovation with practical constraints. The research contributes to organizational readiness theory and practice by illuminating how micro-level perceptions and experiences shape macro-level adoption outcomes.

Current financial LLM agent benchmarks are inadequate. They prioritize task performance while ignoring fundamental safety risks. Threats like hallucinations, temporal misalignment, and adversarial vulnerabilities pose systemic risks in high-stakes financial environments, yet existing evaluation frameworks fail to capture these risks. We take a firm position: traditional benchmarks are insufficient to ensure the reliability of LLM agents in finance. To address this, we analyze existing financial LLM agent benchmarks, finding safety gaps and introducing ten risk-aware evaluation metrics. Through an empirical evaluation of both API-based and open-weight LLM agents, we reveal hidden vulnerabilities that remain undetected by conventional assessments. To move the field forward, we propose the Safety-Aware Evaluation Agent (SAEA), grounded in a three-level evaluation framework that assesses agents at the model level (intrinsic capabilities), workflow level (multi-step process reliability), and system level (integration robustness). Our findings highlight the urgent need to redefine LLM agent evaluation standards by shifting the focus from raw performance to safety, robustness, and real world resilience.

Constitutional AI (CAI) guides LLM behavior using constitutions, but identifying which principles are most effective for model alignment remains an open challenge. We introduce the C3AI framework (\textit{Crafting Constitutions for CAI models}), which serves two key functions: (1) selecting and structuring principles to form effective constitutions before fine-tuning; and (2) evaluating whether fine-tuned CAI models follow these principles in practice. By analyzing principles from AI and psychology, we found that positively framed, behavior-based principles align more closely with human preferences than negatively framed or trait-based principles. In a safety alignment use case, we applied a graph-based principle selection method to refine an existing CAI constitution, improving safety measures while maintaining strong general reasoning capabilities. Interestingly, fine-tuned CAI models performed well on negatively framed principles but struggled with positively framed ones, in contrast to our human alignment results. This highlights a potential gap between principle design and model adherence. Overall, C3AI provides a structured and scalable approach to both crafting and evaluating CAI constitutions.

Training generative AI models requires extensive amounts of data. A common practice is to collect such data through web scraping. Yet, much of what has been and is collected is copyright protected. Its use may be copyright infringement. In the USA, AI developers rely on "fair use" and in Europe, the prevailing view is that the exception for "Text and Data Mining" (TDM) applies. In a recent interdisciplinary tandem-study, we have argued in detail that this is actually not the case because generative AI training fundamentally differs from TDM. In this article, we share our main findings and the implications for both public and corporate research on generative models. We further discuss how the phenomenon of training data memorization leads to copyright issues independently from the "fair use" and TDM exceptions. Finally, we outline how the ISMIR could contribute to the ongoing discussion about fair practices with respect to generative AI that satisfy all stakeholders.

Large language models (LLMs) excel at handling human queries, but they can occasionally generate flawed or unexpected responses. Understanding their internal states is crucial for understanding their successes, diagnosing their failures, and refining their capabilities. Although sparse autoencoders (SAEs) have shown promise for interpreting LLM internal representations, limited research has explored how to better explain SAE features, i.e., understanding the semantic meaning of features learned by SAE. Our theoretical analysis reveals that existing explanation methods suffer from the frequency bias issue, where they emphasize linguistic patterns over semantic concepts, while the latter is more critical to steer LLM behaviors. To address this, we propose using a fixed vocabulary set for feature interpretations and designing a mutual information-based objective, aiming to better capture the semantic meaning behind these features. We further propose two runtime steering strategies that adjust the learned feature activations based on their corresponding explanations. Empirical results show that, compared to baselines, our method provides more discourse-level explanations and effectively steers LLM behaviors to defend against jailbreak attacks. These findings highlight the value of explanations for steering LLM behaviors in downstream applications. We will release our code and data once accepted.

Assessing whether AI-generated images are substantially similar to copyrighted works is a crucial step in resolving copyright disputes. In this paper, we propose CopyJudge, an automated copyright infringement identification framework that leverages large vision-language models (LVLMs) to simulate practical court processes for determining substantial similarity between copyrighted images and those generated by text-to-image diffusion models. Specifically, we employ an abstraction-filtration-comparison test framework with multi-LVLM debate to assess the likelihood of infringement and provide detailed judgment rationales. Based on the judgments, we further introduce a general LVLM-based mitigation strategy that automatically optimizes infringing prompts by avoiding sensitive expressions while preserving the non-infringing content. Besides, our approach can be enhanced by exploring non-infringing noise vectors within the diffusion latent space via reinforcement learning, even without modifying the original prompts. Experimental results show that our identification method achieves comparable state-of-the-art performance, while offering superior generalization and interpretability across various forms of infringement, and that our mitigation method could more effectively mitigate memorization and IP infringement without losing non-infringing expressions.

Spending time on dating apps, I know from experience, can make you a little paranoid. When you swipe and swipe and nothing's working out, it could be that you've had bad luck. It could be that you're too picky. It could be--oh God--that you simply don't pull like you thought you did. But sometimes, whether out of self-protection or righteous skepticism of corporate motives, you might think: Maybe the nameless faces who created this product are conspiring against me to turn a profit--meddling in my dating life so that I'll spend the rest of my days alone, paying for any feature that gives me a shred of hope.

On the Friday after Christmas, OpenAI published a blog post titled "Why OpenAI's structure must evolve to advance our mission." In it, the company detailed a plan to reorganize its for-profit arm into a public benefit corporation (PBC). In the weeks since that announcement, I've spoken to some of the country's leading corporate law experts to gain a better understanding of OpenAI's plan, and, more importantly, what it might mean for its mission to build safe artificial general intelligence (AGI). "Public benefit corporations are a relatively recent addition to the universe of business entity types," says Jens Dammann, professor of corporate law at the University of Texas School of Law. Depending on who you ask, you may get a different history of PBCs, but in the dominant narrative, they came out of a certification program created by a nonprofit called B Lab. Companies that complete a self-assessment and pay an annual fee to B Lab can carry the B Lab logo on their products and websites and call themselves B-Corps.

Despite all beliefs to the contrary, the billionaires who have been seen in President Donald Trump's orbit since he won the presidency for a second time last November are not mere sycophants to his regime. Former Washington Post political cartoonist Ann Telnaes should know. Last month, Telnaes quit her job after her editor refused to publish what turned out to be her last cartoon for the newspaper. In it, Telnaes drew Amazon and Washington Post owner Jeff Bezos, Los Angeles Times owner Patrick Soon-Shiong, OpenAI billionaire Sam Altman, Meta's Mark Zuckerberg, and Mickey Mouse (representing media giant Disney/American Broadcasting Company) either kneeling or bowing face down in front of a statue of the president. In explaining her decision to resign from the Post, Telnaes wrote, "Owners of such press organizations are responsible for safeguarding that free press – and trying to get in the good graces of an autocrat-in-waiting will only result in undermining that free press."

Artificial intelligences picking up sexist and racist biases is a well-known and persistent problem, but researchers are now turning this to their advantage to analyse social attitudes through history. Training AI models on novels from a certain decade can instil them with the prejudices of that era, offering a new way to study how cultural biases have evolved over time. Large language models (LLMs) such as ChatGPT learn by analysing large collections of text.