AI Flaws in Amazon Bedrock, LangSmith, and SGLang Enable Data Exfiltration and RCE
Security researchers disclosed critical flaws in Amazon Bedrock, LangSmith, and SGLang that enable data exfiltration and remote code execution through DNS tunneling, token theft, and unsafe deserialization. Amazon Bedrock's sandbox mode allows outbound DNS queries despite network isolation claims (CVSS 7.5), LangSmith suffers from URL parameter injection enabling account takeover (CVE-2026-25750, CVSS 8.5), and SGLang contains unauthenticated RCE vulnerabilities via unsafe pickle deserialization (CVE-2026-3059/3060, CVSS 9.8). Organizations are urged to migrate to VPC mode, implement DNS firewalls, audit IAM permissions, and patch immediately.
AI summary of the linked source. Not reviewed by an editor. How this works
What helps against this
How we list vendors
- Privileged access management (PAM)
Admin accounts stop being one stolen password away from control of everything.
- Cloud security posture management (CSPM and CNAPP)
Misconfigured cloud storage and over-permissive roles get caught before they leak.
- AI application security
Prompt injection, data leaks through AI tools and unapproved AI use get caught.
Indicators of Compromise
Extracted by AI from the linked source. Not reviewed by an editor. How this works
- cve — CVE-2026-25750
- cve — CVE-2026-3059
- cve — CVE-2026-3060
- malware — DNS tunneling C2 channel
- domain — smith.langchain[.]com