AWS Launches Continuum and Context to Fix Agent Blind Spots

#ai #programming #tech #product

AWS launched Continuum and Context to fix AI agent security and context gaps. Both services automate vulnerability handling and knowledge graph construction.

At the AWS Summit in New York, Amazon launched Continuum and Context to fix AI agents' lack of business context and security. Both services target the widening gap between agent speed and enterprise readiness.

Key facts

AWS Continuum covers full vulnerability lifecycle: detection to remediation.
Continuum uses frontier models like Anthropic's Claude Mythos.
AWS Context builds a knowledge graph from corporate data.
Continuum starts in learning mode, optional enforcement mode.
Both services initially available to select pilot customers.

AWS unveiled two new services at its New York summit designed to address the most common failures of AI agents in production: security vulnerabilities and lack of business context. According to The Decoder

AWS Continuum: Automated Vulnerability Lifecycle

AWS Continuum covers the full lifecycle of code vulnerabilities — detection, prioritization, validation, and remediation. The service scans for new vulnerabilities on its own, ranks findings by business context (e.g., is the affected component reachable? is it in production?), then validates them by attempting an attack in an isolated test environment. Only after validation does it suggest countermeasures like network config changes, permission adjustments, or code patches.

Continuum picks different frontier models depending on the task, including specialized security models like Anthropic's Claude Mythos. It starts in a learning mode requiring human sign-off, with an option to switch to enforcement mode where it applies fixes autonomously. A companion threat modeling tool generates attack scenarios from design documents or source code. Initially available only to select pilot customers.

AWS Context: Knowledge Graph for Agents

AWS Context builds a shared knowledge graph from corporate data, giving AI agents the business context they currently lack. The service ingests internal documents, databases, and APIs to create a structured representation of an organization's domain, policies, and workflows. Agents query this graph before making decisions, reducing the risk of hallucinated or irrelevant outputs.

The Bigger Picture

Both services address a structural problem: agents that write code fast but get things wrong too often. AWS's move mirrors a broader industry trend — enterprise customers are demanding guardrails as they deploy agents beyond demos. Amazon's $4B+ investment in Anthropic gives it access to models like Claude Mythos, which AWS claims can spot vulnerabilities faster than defenders can respond.

But the gap between agent speed and enterprise readiness remains wide. Continuum and Context are patches, not a fundamental fix — they layer security and context onto agents that still lack robust reasoning. The real test will be whether these services can keep pace as agents become more autonomous.

What to Watch

Watch for Continuum's general availability date and pricing, expected within the next quarter. Key metric: how many pilot customers adopt enforcement mode over learning mode, signaling enterprise trust in automated remediation. Also monitor whether AWS integrates Context into Bedrock AgentCore, which would extend the knowledge graph to third-party agents.