Artificial Insecurity: how AI tools compromise confidentiality

European policymakers are increasingly grappling with the inherent security vulnerabilities within artificial intelligence (AI) tools, a critical issue impacting data confidentiality and information integrity. Recent analyses highlight that these sophisticated technologies, despite their promise, are often built upon insecure foundations, posing significant risks that demand urgent attention within the digital policy landscape. These AI systems can inadvertently expose sensitive user data and even facilitate malicious hacking attempts due to underlying dodgy security practices. This extends to undermining the integrity of information and creating critical vulnerabilities within digital supply chains, raising profound concerns for businesses and individuals alike. The implications of these security shortcomings are far-reaching, affecting the confidentiality of personal data, the reliability of information, and the overall availability of digital systems. A human rights-respecting approach to AI development and deployment is being advocated as a potential solution to mitigate these grave consequences.