Amazon convened an emergency meeting after outages caused by code changes using generative AI.

Amazon faced a series of technical issues last week.

It is believed that the outages may have been caused by the use of generative artificial intelligence (AI) in the code development process. Internal emails and employee meetings confirmed this hypothesis, but the company emphasizes that the errors are not related to the AWS cloud division.

What happened
Date | Event | Explanation
---|---|---
Friday | First delay: users could not complete purchases, view personal data or see prices. An internal email reported widespread software updates; later clarification – issues were linked to AI‑assisted development.
Tuesday | Four incidents affecting critical infrastructure. Senior VP Dave Treadwell admitted that stability “was not flawless.”
Thursday | Around six o’clock the Amazon website and app became unavailable for some users. Initially it was said to be due to updates; later it was attributed to insufficient control over AI‑generated code changes.

How the company is responding
1. Regular meetings

- Weekly incident specialist meetings.
- Agenda: site availability issues, mobile app problems, and AI usage questions.

2. Control of AI‑generated changes

- Strengthening the AI‑initiated code review process.
- Temporary additional safeguards are planned.

3. Separating AWS issues

- Confirmed that outages do not involve the AWS cloud division.
- In December, problems arose after implementing changes suggested by the AI agent Kiro; the company again focused on human factors.

Conclusion
- Last week’s outages may have been caused by generative AI usage.
- Internal communications (emails, meetings) confirmed this hypothesis but clarified that only one case was affected.
- Amazon intends to tighten control over the AI‑assisted code development process and will continue weekly incident discussions.