Amazon Cloud AI Tools Cause Small Outages in December 2025

Amazon Cloud (AWS) had at least two small outages in December 2025. Reports say internal AI tools caused these, not user error.
By Newsroom | Technology, Cloud Computing |

Amazon Web Services (AWS), a leading provider of cloud computing, has faced at least two significant service disruptions in recent months that are reportedly linked to the use of internal artificial intelligence (AI) tools. While Amazon states these were minor incidents, the implications for the stability of large-scale cloud infrastructure are under scrutiny. The company asserts that user error, not AI, was the cause, yet reports from individuals familiar with the matter suggest that AI tools were granted extensive permissions, leading to the disruptions.

Context of the Incidents

In December 2025, at least two separate incidents impacted AWS services. One of these, described as an "extremely limited event," affected a single service in one of two regions in mainland China. According to reports, an agentic tool, capable of independent action, initiated a "delete and recreate the environment" command. This occurred despite the AI tools being granted the same permissions as human engineers and their output not undergoing the usual approval processes.

Read More: Persona age verification system exposed user data including ID documents and selfies

Amazon’s cloud ‘hit by two outages caused by AI tools last year’ - 1

An earlier, larger disruption occurred in October 2025, causing a 15-hour global outage. This event affected over 1,000 companies and disrupted services like Reddit, Roblox, and Snapchat. While Amazon has publicly stated that the December incidents were brief and attributed them to user error, particularly misconfigured access controls, information obtained by the Financial Times from unnamed sources suggests a different narrative. These sources claim that the AI tools were given broad access and their actions were not sufficiently reviewed.

Nature of the AI Tools and Outages

The AI tools in question are described as internal systems designed to manage AWS infrastructure. Reports indicate that these agentic AI tools have the capacity to perform autonomous actions for users, such as deleting and recreating environments.

Amazon’s cloud ‘hit by two outages caused by AI tools last year’ - 2

  • December Incidents:

  • At least two outages reported.

  • One incident involved an AI tool attempting to "delete and recreate the environment."

  • The cause was attributed by Amazon to "user error" and "misconfigured access controls."

  • Sources suggest AI tools were granted elevated permissions without standard oversight.

  • October Outage:

  • A major, 15-hour global disruption affecting numerous companies and services.

  • While the primary cause of this specific outage is detailed in other reports, the broader context of AI's role in infrastructure management is being examined.

Differing Perspectives on Causation

Amazon's official stance attributes the recent disruptions to user error, specifically citing "misconfigured access controls" rather than any fault of the AI systems themselves. A spokesperson stated that the events were brief, extremely limited, and did not impact core AWS services like compute, storage, or AI technologies.

Read More: OpenAI plans screenless AI device with Jony Ive for 2026 launch

However, reporting from the Financial Times, based on four unnamed individuals familiar with the matter, presents a contrasting view. These sources suggest that the AI tools were involved and were granted permissions akin to those of human engineers. They also indicated that the AI's actions were not subjected to the same level of approval that would typically be applied to human-generated changes.

Amazon’s cloud ‘hit by two outages caused by AI tools last year’ - 3

"The FT's sources suggest incorrect permissions were to blame, with the AI tools given the same permissions as human workers and its output not given the same approval as would usually be the case with human workers." - TechRadar

Amazon confirmed plans for significant job cuts in January 2026, following layoffs in October 2025. While CEO Andy Jassy reportedly framed these cuts as being related to company culture rather than AI replacement, the timing of the outages raises questions about the integration of AI in operational roles. Amazon has stated that there is no evidence to suggest AI technology leads to more errors than human engineers.

Expert Insights on AI and System Stability

Cybersecurity experts highlight the inherent complexities of advanced AI systems. Michał Woźniak, a cybersecurity expert, suggests that completely preventing errors from internal AI agents could be nearly impossible due to the unpredictable choices AI systems can make and their extreme complexity. This complexity means that even with safeguards, AI systems might behave in ways that are difficult for human operators to anticipate.

Read More: HIMSS26 Symposium in March 2026 to Discuss Native American Health Tech

Amazon’s cloud ‘hit by two outages caused by AI tools last year’ - 4

"It would be nearly impossible for Amazon to completely prevent internal AI agents from making errors in future, because AI systems make unexpected choices and are extremely complex." - Michał Woźniak, cybersecurity expert

These AI-linked outages prompt considerations regarding the balance between the efficiency gains offered by autonomous AI tooling and the necessity of robust risk controls in critical infrastructure.

Findings and Implications

The reports of AWS outages linked to internal AI tools, despite Amazon's official attributions, introduce a level of uncertainty regarding the direct causal role of AI.

  • Evidence of AI Involvement: Multiple sources and reports suggest that internal AI tools were active during the incidents and were granted significant operational permissions.

  • Conflicting Causality: Amazon maintains that user error was the primary cause, while other reports point to the AI tools' actions and permissions as contributing factors.

  • Systemic Complexity: The inherent complexity and potential unpredictability of AI systems are cited as challenges in preventing future errors.

  • Risk Management Concerns: The incidents highlight the ongoing need for companies to effectively balance the benefits of AI automation with rigorous risk management protocols.

The long-term implications for cloud infrastructure stability and the adoption of AI in operational roles remain a subject of ongoing observation. Further transparency from AWS regarding the specific nature of these AI tools and their operational parameters may be necessary to fully assess the risks.

Sources

Frequently Asked Questions

Q: What happened with Amazon Cloud (AWS) in December 2025?
Amazon Web Services (AWS) had at least two small service disruptions in December 2025. One incident involved an AI tool trying to delete and recreate its environment.
Q: Why did the Amazon Cloud (AWS) outages happen in December 2025?
Amazon says user error and wrong settings caused the outages. But some reports say internal AI tools, given high access, were the reason.
Q: Who was affected by the Amazon Cloud (AWS) AI tool incidents in December 2025?
The incidents were described as very small and affecting only one service in a limited area. Amazon states core services like compute and storage were not impacted.
Q: What is Amazon's official reason for the December 2025 AWS outages?
Amazon officially stated that the outages were caused by user error, specifically misconfigured access controls, and were very brief and limited.
Q: What do other reports say caused the December 2025 AWS outages?
Reports from sources familiar with the matter suggest that internal AI tools were given the same high permissions as human workers and their actions were not checked properly, leading to the disruptions.