Artificial Intelligence & SaaS
Strands Agents Tools 0.8.6 Fixes Secret-Masking Bypass
Strands Agents Tools 0.8.6 fixes a secret-masking bypass. Upgrade, find loaded environment tools, rotate credentials, and retain evidence.
AWS Remediates MCP Server run_script Sandbox Validation
AWS remediated MCP Server run_script sandbox validation service-side. Confirm service use and access governance, document status, and retain evidence.
AWS Directs Multi Model Server Users to Migrate After Two Flaws
AWS says every Multi Model Server release is affected and no fix is planned. Restrict management access, migrate, validate workloads, and retain evidence.
AWS Strands Agents SDK Updates Isolate A2A Conversation State
AWS fixed A2A conversation-state isolation in Strands Agents SDKs. Upgrade Python and TypeScript packages, validate tenancy, and retain evidence.
AWS Fixes SageMaker SDK Default-Bucket Ownership Flaw
AWS fixed SageMaker SDK default-bucket ownership checks. Upgrade v2 and v3, verify S3 ownership, validate jobs, and retain IAM evidence.
AWS Kiro Crew 0.7.2 Fixes WhatsApp Cross-Session Authorization Flaw
AWS fixed a Kiro Crew WhatsApp authorization flaw in 0.7.2. Check non-default channel settings, update or disable the channel, and retain evidence.
AWS databases-on-aws 1.7.1 Blocks Agent-Assisted Command Injection
Amazon Web Services issued 2026-130-AWS. Match affected deployments, apply supported fixes, validate operations, and retain evidence.
AWS Fixes SageMaker Studio Space Startup Command Injection
Amazon Web Services (AWS) issued 2026-125-AWS. Match affected deployments, apply supported fixes, validate operations, and retain evidence.
GitLab AI Gateway Updates Fix Critical CVE-2026-90970
Self-hosted GitLab AI Gateways need branch-specific fixes for authenticated prompt-template sandbox escape through crafted flows. Identify who hosts the gateway: updating the main GitLab application does not prove this separate component is fixed.
LightRAG Security Update Addresses 5 Vulnerabilities
LightRAG 1.5.5 addresses API disclosure, plaintext-password timing, login throttling, Markdown-image SSRF and stored WebUI script injection. Update backend and frontend components, then check the actual authentication, ingestion and rate-limit configuration.