In November, Amazon’s leadership issued an internal memo encouraging employees to utilize its proprietary code generation tool, Kiro, instead of third-party options from competitors.
“While we continue to support existing tools currently in use, we do not plan to endorse additional third-party, AI development tools,” read the memo, as Reuters reported at the time. “As part of our builder community, you all play a vital role in shaping these products, and we rely on your feedback to enhance them rapidly.”
This directive was surprising, given the substantial investments—amounting to tens of billions of dollars—Amazon has made in competitors like Anthropic and OpenAI. These tech firms are currently engaged in a fierce competition to dominate the burgeoning AI coding sector, while Amazon’s Kiro has lagged far behind.
Fast forward six months, and Amazon’s stance has shifted dramatically. As Business Insider reports, the company has officially decided to allow its employees access to OpenAI’s Codex and Anthropic’s Claude, resulting from increasing employee requests.
This decision underscores the intense pressure AI companies face to maintain their competitive advantages—and, importantly, to avoid financial setbacks. For Amazon, which has established strong partnerships within the AI landscape as part of its expansive cloud strategy, this turn of events is particularly awkward.
Moreover, Amazon’s own attempts to enhance its AI coding tools have not gone smoothly, with the company acknowledging that recent outages were linked to inadequately executed AI-generated code.
In a note shared with staff that was obtained by BI, Jim Haughwout, Vice President of Amazon Software Builder Experience, announced that Claude Code would be made available, with Codex to follow shortly afterward.
Although this move might appear to be a concession, it’s worth noting that both coding tools will operate on Amazon’s Bedrock, a fully managed service within Amazon Web Services designed to provide secure access to advanced AI models. However, the gesture does suggest an acknowledgment that the company’s flagship tool isn’t faring well in comparison to its rivals.
“To empower you to create more value for our customers, we are broadening the available AI tools,” Haughwout informed employees.
Earlier this year, software developers within the company had expressed frustration with the restrictions placed on the use of Claude Code, as highlighted in the November memo. Many found it contradictory to promote Claude Code via AWS Bedrock while being unable to utilize it for their work.
“Customers will want to know why they should trust or utilize a tool that we have not approved for internal use,” one employee remarked in feedback obtained by BI.
Despite the optics of this significant policy change, an Amazon spokesperson stated that teams are still “primarily using” Kiro, claiming that 83 percent of engineers continue to rely on it.
More on Amazon: Amazon Admits Extensive AI Use Is Wreaking Havoc on Its Core Business
In this article, we explore Amazon’s evolving approach to AI coding tools. Initially resisting reliance on third-party options, the company has since reversed course, now allowing access to leading AI solutions. This shift reflects both internal pressures and the competitive landscape of the tech industry.
As Amazon adapts to the rapid advancements in AI, the decision highlights the ongoing challenge tech companies face: balancing internal development with the necessities of modern innovation. This evolution may signal a substantial change in the future of coding practices at Amazon.
