Full Summary
This Friday morning, both Techerati and the National Cyber Security Centre confirm that AI is dramatically escalating cybersecurity risks, making attacks easier, faster, and cheaper. Attackers are now refining prompts over extended conversations, exposing vulnerabilities previously missed by single-turn benchmarks. Cisco's research shows multi-turn attack success rates jumped from a range of 2.19%-64.91% to 7.89%-88.30% across 15 models. Google Cloud is responding by introducing new AI tools to combat these emerging threats, as attackers compress vulnerability research and malware development timelines from weeks to hours. Similarly, the EC-Council, with input from Citi and Microsoft, launched its ADG framework to help organizations securely adopt and govern AI at scale, a crucial step given only 1% of leaders believe their AI governance is mature. The AI Journal highlights a critical shift: traditional application security can no longer keep up with AI-generated code, which is created and deployed at high speeds without human review. This means security needs to move beyond static code analysis. The Bank of Italy is engaging with AI firms to address security risks for banks, while CrowdStrike has expanded Project QuiltWorks to link AI-driven security data with insurance risk models, aiming to quantify and transfer financial risk from frontier AI threats. For you, this means your personal data and financial transactions face unprecedented risks from AI-powered attacks like deepfakes. It's critical to regularly update all your devices and software, as highlighted by The American Legion, and be wary of AI agents in financial services like those offered by Robinhood, which raise concerns about accountability if bad advice leads to losses.