Amazon Reverses AI Policy, Grants Engineers Access to Codex and Claude

In November, Amazon’s leadership sent a memo telling staff to rely exclusively on the company’s own code‑generation system, Kiro, and to reject any new external AI development tools. The directive emphasized that while existing tools would remain supported, no additional third‑party solutions would be approved.

Six months later, Vice President Jim Haughwout announced a reversal. He told employees that the firm is expanding the “agentic AI tools” they can use, specifically naming Claude Code and, shortly thereafter, Codex. Both models will run on Amazon Bedrock, the cloud service that provides secure, managed access to cutting‑edge AI.

The shift follows growing frustration among developers who found Kiro lagging behind competitors and were blocked from using Claude Code internally despite its promotion to customers. Outages linked to AI‑generated code further eroded confidence in the in‑house tool.

Amazon’s spokesperson noted that 83 % of engineers still rely on Kiro as their primary coding assistant, indicating that the majority of the workforce has not yet migrated to the new options. Nonetheless, the policy change signals that the company acknowledges Kiro’s shortcomings relative to market leaders.

Key facts - November memo: no support for additional third‑party AI tools. - VP Haughwout: “We are expanding the agentic AI tools available to you.” - 83 % of engineers primarily use Kiro, according to internal data.

What it means Opening Codex and Claude gives Amazon engineers access to models that have demonstrated higher accuracy and broader language support than Kiro. The move may improve developer productivity and reduce the risk of AI‑induced outages. It also aligns Amazon’s internal tooling with the services it sells to external customers via Bedrock, removing the inconsistency of promoting a tool it does not use itself.

The decision underscores the pressure on large tech firms to stay competitive in the fast‑moving AI coding market. As Amazon integrates these external models, watch for metrics on adoption rates, changes in outage frequency, and any further adjustments to its internal AI strategy.