Artificial intelligence news.
Straight from the original announcements.

Tamnoon Builds Tami, an AI Agent for Cloud Security Fixes, on Amazon Bedrock

October 14, 2026, 15:53By the AI News Bites Editorial TeamSource: AWS
Tamnoon Builds Tami, an AI Agent for Cloud Security Fixes, on Amazon Bedrock
Image: AWS

Tamnoon has built Tami, an agentic cloud remediation orchestrator on Amazon Bedrock that investigates security findings, judges whether a fix is safe and orchestrates the change. According to AWS, the approach drove an over 90 percent reduction in critical cloud exposure in 90 days with zero production issues.

Key points
  • Tamnoon built Tami, an agentic cloud remediation orchestrator, on Amazon Bedrock.
  • According to AWS, the approach cut critical cloud exposure by over 90 percent in 90 days with zero production issues.
  • Four agents handle investigation, safety scoring, remediation planning and code generation, with guardrails and a reviewer agent.
  • Each finding gets a Remediation Confidence Indicator of SAFE, RISKY or AWAITING DATA.
  • Inference runs inside Tamnoon's own AWS environment, and sensitive customer data stays within the trust boundary, according to AWS.

The problem: findings pile up faster than teams can fix them

AWS describes a familiar gap in cloud security. Cloud-native application protection platforms (CNAPP), cloud providers and posture scanners generate more findings than any team can triage, and risk builds up between detection and repair. According to the post, most findings sit untouched not because they are hard to fix, but because someone must investigate context, identify the owner and confirm the fix will not break production.

The post lists four challenges: alert volume that overwhelms headcount, manual and repetitive investigation, fear of production impact that blocks automation, and reactive prioritization that works through alerts linearly instead of addressing root causes. A security engineer can spend hours on a single finding, and it often traces back to the same root cause as dozens of other alerts.

How Tami works: four specialized agents

Tamnoon is a cloud security remediation platform that works alongside CNAPPs. It ingests findings from CNAPP and Cloud Security Posture Management (CSPM) providers and other tools, normalizes them into a unified schema, enriches them with asset and exposure context and prioritizes them. An agentic layer then takes over, running in an event-driven, serverless architecture on AWS.

The platform delegates work to four agents:

  • Investigation agent: draws on training from millions of real customer cloud environments, filters context through an LLM-as-a-judge validation layer and produces read-only investigation steps and a complete investigation plan.
  • Safety-scoring agent: builds an evidence chain per asset and classifies each remediation as SAFE, RISKY or AWAITING DATA, with a deliberately conservative bias. When in doubt, it downgrades.
  • Remediation agent: activates only for findings deemed safe and produces a plan in which every step follows a backup, remediate and verify structure, with rollback guidance.
  • Code-generation agent: turns plan steps into executable remediation after an input guardrail screens for relevance and prompt injection. A separate reviewer agent must approve every artifact.

For each finding, Tami assigns a Remediation Confidence Indicator (RCI) of SAFE, RISKY or AWAITING DATA. That rating tells teams what can be automated and what needs human review first.

Why Amazon Bedrock and how it connects to existing tools

Tamnoon's agents do different kinds of work: deep investigative reasoning, high-volume validation and code review. According to the post, matching each job to the right model without a separate integration for each was a core requirement. Through a single API on Amazon Bedrock, each agent can call the foundation model best suited to its task, and models can be swapped or upgraded without re-plumbing the architecture.

Two other factors are cited. Amazon Bedrock Cross-Region inference routes requests to the optimal AWS Region within a defined geographic boundary, so Tamnoon avoids writing its own traffic distribution logic. On security, inference runs inside Tamnoon's own AWS environment under AWS Identity and Access Management (IAM)-scoped permissions, and sensitive customer data never leaves the trust boundary, according to AWS.

Tamnoon sits downstream of existing detection tools. The post names CNAPPs such as Wiz, CrowdStrike, Palo Alto Cortex, Upwind, SentinelOne, Orca and Cyera, along with AWS Security Hub and Amazon GuardDuty. Findings are normalized so a misconfiguration looks the same whether it came from Wiz or GuardDuty.

The AWS architecture behind the platform

Tamnoon runs entirely within its own AWS accounts, isolated per environment. Users enter through Amazon API Gateway, with encryption managed by AWS Certificate Manager and sign-in handled by Amazon Cognito. Everything behind that entry point runs in private networks across multiple Availability Zones.

Amazon Elastic Kubernetes Service (Amazon EKS) runs the always-on services behind the dashboard, while AWS Lambda handles the event-driven work of collecting findings, analyzing them and executing remediations. Amazon EventBridge, Amazon Simple Queue Service (Amazon SQS) and AWS Step Functions coordinate these steps so large volumes are processed reliably and in parallel, with no single bad record stalling the rest.

The AI layer runs on Amazon Bedrock. Amazon Bedrock AgentCore gives the agents a governed way to take action instead of open-ended access, and Amazon Bedrock Guardrails review both the request and the generated code before anything moves forward.

What it means for security teams

The case study points to a model in which detection tools stay in place and a separate layer handles the investigation, safety checks and fixes that follow. The conservative SAFE, RISKY and AWAITING DATA rating is meant to address the fear of production impact that, per the post, keeps teams from automating remediation. The reported results, an over 90 percent reduction in critical cloud exposure in 90 days with zero production issues, are AWS's account of the approach and come from the post itself.

Related news

For developers
For developersNovember 13, 2026, 17:14

Open Agentic Web Challenge offers $13,000 for interoperable AI agents

A global group of technology, research, standards and civil society organizations has launched the Open Agentic Web Challenge, offering $13,000 for open and…

AWS upgrades AgentCore runtime with faster starts and stronger isolation
Products and toolsNovember 12, 2026, 18:59

AWS upgrades AgentCore runtime with faster starts and stronger isolation

AWS has announced updates to the Amazon Bedrock AgentCore runtime, giving AI agent sessions faster startup, more compute capacity and expanded isolation for data. The…

Red Hat OpenShift Lightspeed 2 adds live and autonomous troubleshooting
Products and toolsNovember 12, 2026

Red Hat OpenShift Lightspeed 2 adds live and autonomous troubleshooting

Red Hat OpenShift Lightspeed 2 can now investigate OpenShift cluster problems directly, either through a live troubleshooting chat or an autonomous agentic run. The…