Bedrock flaw exposes deeper cloud AI risks
Security researchers have uncovered a vulnerability in Amazon Web Services Bedrock’s code interpreter environment, raising concerns over the robustness of isolation safeguards in generative AI systems and prompting renewed scrutiny of cloud-based development tools used by enterprises. The flaw, identified in Bedrock’s sandboxed execution layer, centres on weaknesses in DNS handling that could allow data exfiltration from supposedly isolated environments. Analysts say the issue highlights structural challenges […]