Data quality is crucial for generating accurate results in AI models. Gen AI can generate quality synthetic data for testing models, but human oversight is important. Generative AI models can behave unpredictably and may pose challenges in understanding and explaining their decisions. Gen AI is particularly important in the banking and finance sectors, where it can generate realistic synthetic customer data for testing models and improving risk assessment. Banks like Capital One and JPMorgan Chase use Gen AI to strengthen their fraud detection systems. Moving quickly to scale generative AI across organizations in the banking sector can result in significant benefits. However, careful model governance, oversight, explainability, and integration into existing systems are necessary due to the highly regulated nature of the industry [c249234a].
Scientists at CyLab have developed a taxonomy of privacy risks associated with AI. The researchers examined 321 recorded AI privacy events and used Daniel J. Solove's 2006 publication 'A Taxonomy of Privacy' as a baseline. They identified 12 high-level privacy risks that are newly created or exacerbated by AI technologies, including physiognomy and data breaches due to insecurity [2dc1761c].
This taxonomy of AI privacy risks highlights the growing concerns surrounding the use of AI and the potential risks to privacy. As AI technologies continue to evolve, it is crucial to understand and address these risks to ensure responsible and secure use of the technology. The taxonomy provides a framework for identifying and categorizing privacy risks, which can inform the development of regulatory measures and best practices for AI systems. By addressing these risks, stakeholders can work towards mitigating potential privacy breaches and safeguarding sensitive data [2dc1761c].
In response to the increasing concerns about AI risks, the Responsible Investment Association of Australasia (RIAA) has launched the Artificial Intelligence and Human Rights Investor Toolkit. The toolkit aims to help advisers address the risks presented by AI and provides case studies, methodologies for understanding risks, and guidance for investor engagement. It was developed in response to concerns raised at RIAA's national conference about digital technology issues and the increasing use of AI across industries. The toolkit aims to help firms implement and deploy AI ethically and responsibly. RIAA has over 500 members representing US$29 trillion in assets under management [1463b03c].
Investors are warned about the risks of artificial intelligence (AI) abuse, including discrimination, loss of privacy, and distorted data. The RIAA toolkit advises on financial and human rights risks associated with AI, including the amplification of negative content by algorithms on social media platforms, cyber-attacks resulting in privacy violations, bias and discrimination, and the creation of sexually explicit deepfake imagery. The RIAA recommends a method called 'stewardship' to reduce risk, involving direct communication with company bosses to address concerns or divest from unsafe investments [c0a3fae7].
The article 'The AI-Powered Metaverse: Profound Privacy Risks and Dangers' discusses the privacy risks and dangers associated with the AI-powered metaverse. The author, Alvin Wang Graylin, expresses concern about the potential for large corporations or state actors to track and control users' experiences in virtual and augmented worlds. These platforms have the capability to monitor users' movements, actions, emotions, and even physiological responses. The author argues that existing online privacy protections are insufficient for immersive worlds and calls for proactive regulation to protect against privacy violations. The article emphasizes the need to balance the value of tracking behaviors and emotions with the privacy risks involved. Stored data poses a greater danger as it can be processed by AI to create behavioral and emotional models that predict users' actions and feelings. The author suggests that regulation could help build trust among the public and ensure responsible use of AI in the metaverse [a95d2ccf].
The article 'Privageddon: The Convergence of Privacy, AI, and Digital Currencies' discusses the Privageddon event, a week-long virtual assembly focused on privacy and artificial intelligence (AI) that transitions into a physical meet-up in Milan. The event explores the impact of AI on professional realms, the evolution of digital payments, and the laws framing the digital world. It addresses the challenges and controversies surrounding legislation, including concerns about privacy, security, and the need for targeted legal frameworks. The article also highlights the advantages and disadvantages of the convergence of AI and digital currencies, such as enhanced surveillance capabilities and risks to privacy. It provides links to resources on the AI Act and GDPR for further information [c0ae3c39].
The International Swaps and Derivatives Association (ISDA) has published a report on AI in derivatives, emphasizing the importance of finding a balance between leveraging AI opportunities and managing risks. The report, authored by Tom Reynolds and Katherine Arden, highlights the need for developing frameworks to mitigate risk in AI-driven derivatives markets. The authors suggest that the derivatives industry is at a critical moment where decisions made now will shape the future of AI in the sector. They urge market participants to consider the ethical implications of AI and ensure transparency and accountability in its use. The report also highlights the potential benefits of AI in improving efficiency, reducing costs, and enhancing risk management in derivatives trading. The authors call for collaboration between market participants, regulators, and technology providers to develop best practices and standards for AI in derivatives [57052688].
GenAI is set to redefine the operational landscapes of the derivatives market, according to a new paper from ISDA Future Leaders in Derivatives (IFLD). The output from Gen AI models is astonishing and often indistinguishable from what a human might create. Research from McKinsey suggests Gen AI could add as much as $4.4 trillion to global corporate profits annually. The derivatives market plays a pivotal role in global finance but grapples with operational inefficiencies and regulatory complexities. Gen AI presents both opportunities and challenges in the derivatives industry. It has the potential for new efficiencies and opportunities for innovation, but also inherent risks that could severely impact financial firms deploying the technology. Potential use cases for Gen AI in the market include summarizing long documents, synthesizing thoughts, generating new ideas, and helping lawyers navigate challenges. There is scope for Gen AI in the regulatory space, providing tools for regulators to interpret data and detect anomalies. Unconstrained AI represents the risk of amplifying or exacerbating market shocks. Concentration risk and misinformation are also concerns. Policymakers and regulators are trying to catch up with the rapid development of Gen AI technology. Continuous dialogue, collaboration, and regulatory advancements are important for industry stakeholders. Bill Ackman's Pershing Square sold a $1.05 billion stake. UBS Americas named Rob Karofsky and Chris Knight to LMAX. Nasdaq collaborated with Microsoft on the boardroom experience. 42.5% of fraud attempts detected use AI, according to a report. SimCorp - Axioma was awarded the Best M&A Deal. [cc826d9d] [57052688]
Canadian businesses are embracing generative AI for increased productivity and work quality, but it also poses risks such as data leaks and intellectual property ownership. Organizations need to address these risks through strategic alignment and oversight, policies and procedures, and third-party oversight. KPMG recommends building a cross-functional governance committee, adopting an enterprise approach to AI use cases, updating company policies and processes, and mitigating ethical risks of open source platforms. The National Institute of Standards and Technology and the European Union have released guidelines for AI risk management, and KPMG has developed a Trusted AI Framework based on 10 key pillars. By embedding trust within the AI platform and adopting responsible AI practices, businesses can maximize the benefits of generative AI while minimizing risks. [1061441d]
Sam Clarke, media and entertainment underwriter at Markel, questions the comfort level of underwriters with the intellectual property exposures of generative artificial intelligence (AI) when each jurisdiction would effectively view and handle claims differently. Companies have been using AI for years, with virtual assistants being one example. Clarke emphasizes the need for caution and careful consideration of the risks associated with AI [e501f1f9].
Machine learning models have the potential to violate privacy by memorizing sensitive information from the training data. While complex machine learning models can learn more complex patterns, they also have the risk of overfitting to the data, meaning they make accurate predictions about the training data but perform poorly on new data. Machine learning models have parameters that can be adjusted to affect their performance, and they are trained using data to minimize predictive error. However, due to the large number of parameters, machine learning models can memorize some of the training data, compromising the privacy of individuals. Differential privacy is a method that can protect privacy by limiting how much a machine learning model depends on individual data, but it also limits the performance of the model. The trade-off between inferential learning and privacy concerns raises the question of which is more important in different contexts [1603c546].