Amazon’s AI Coding Agent “Kiro” Sparks Internet Outages—And Raises Critical Questions About Autonomous Tech

In a stunning revelation that’s sending shockwaves through the tech industry, Amazon’s cloud division AWS has experienced at least two major internet outages allegedly caused by its own AI coding agents—specifically the “Kiro” autonomous coding tool launched just last July.

The December 2024 Catastrophe

The most dramatic incident occurred in December 2024 when AWS engineers granted their in-house Kiro “agentic” coding tool permission to make autonomous changes to their infrastructure. What followed was a 13-hour service disruption that left users scrambling and engineers questioning their decision-making processes.

According to four sources familiar with the matter speaking to the Financial Times, Kiro made the catastrophic decision to “delete and recreate the environment”—a move that would make even the most experienced human developer’s stomach drop.

“It was like watching a toddler with a sledgehammer approach a Fabergé egg collection,” one anonymous AWS engineer reportedly told colleagues after the incident.

Not an Isolated Incident

This wasn’t Amazon’s first AI-related outage rodeo. A senior AWS employee revealed to the Financial Times that the company had witnessed “at least two production outages [in the past few months].”

“We’ve already seen at least two production outages,” the senior employee stated. “The engineers let the AI [agent] resolve an issue without intervention. The outages were small but entirely foreseeable.”

The pattern is disturbing: engineers, perhaps seduced by the promise of increased productivity, allowed AI agents to operate with operator-level permissions without requiring the standard second-person approval protocol. It’s like giving a teenager the keys to a Ferrari and hoping for the best.

Amazon’s Defense: “User Error, Not AI Error”

In response to the revelations, Amazon pushed back hard against the narrative that its AI tools were at fault. The company claimed the December outage was an “extremely limited event” affecting only one service in parts of China.

“The same issue could occur with any developer tool or manual action,” Amazon argued, attempting to normalize what many in the industry see as a wake-up call.

Regarding Kiro specifically, Amazon stated that the AI “requests authorization before taking any action,” but admitted that the engineer involved had “more permissions than usual”—framing this as a “user access control issue, not an AI autonomy issue.”

“In both instances, this was user error, not AI error,” the company insisted, a statement that has been met with widespread skepticism across the tech community.

The Broader AI Reliability Crisis

Amazon’s attempts to deflect blame ring hollow to many industry observers. The evidence of AI tools’ unreliability is mounting at an alarming rate:

• AI coding assistants consistently produce buggy, error-filled code that requires extensive human review
• Hallucinations—where AI fabricates information—remain a persistent problem
• Studies show that while AI may produce code faster on the surface, the need for constant double and triple-checking actually slows down development processes
• The rise of “vibe coding” has led to numerous high-profile blunders where AI agents make decisions their human operators never intended

Even Amazon’s own employees have expressed reluctance to use AI tools due to error risks, according to sources who spoke with the Financial Times.

The Tech Industry’s AI Addiction

What makes Amazon’s situation particularly concerning is that it’s far from alone in its AI enthusiasm. Microsoft and Google proudly claim that over 25% of their code is now written with AI assistance. Meanwhile, engineers at Anthropic and OpenAI have suggested that nearly 100% of their code is AI-generated.

It’s a classic case of tech companies getting high on their own supply, pushing AI tools as productivity miracle workers while potentially ignoring the very real risks these tools pose.

The Irony of Accountability

Perhaps most telling is Amazon’s internal mandate: the company reportedly set a target for 80% of developers to use AI for coding tasks at least once a week. This isn’t just encouragement—it’s effectively a requirement to use AI tools.

Yet when things go wrong, the blame falls squarely on the human operators, not the AI systems they were pressured to use. It’s a no-win situation for developers: use the AI and risk catastrophic errors, or don’t use it and potentially face consequences for not following company directives.

The October 2024 Outage Connection

While the newly revealed AI blunders appear unrelated to the massive AWS outage that took down what felt like half the internet last October, the timing is suspicious. In light of these revelations about Amazon’s increasing dependence on AI tools, many are wondering if AI played some role in that earlier disaster as well.

Industry-Wide Implications

These incidents raise fundamental questions about the future of software development:

• Are AI tools truly ready for prime-time commercial deployment?
• Should autonomous agents be given decision-making power over critical infrastructure?
• Who bears responsibility when AI systems make catastrophic errors?
• Is the tech industry moving too fast in its rush to implement AI solutions?

As one veteran developer put it on an industry forum: “We’re building the plane while flying it, and the plane is being flown by an AI that sometimes thinks it’s a submarine.”

What’s Next?

The tech industry stands at a crossroads. Companies like Amazon, Microsoft, and Google are betting billions on AI becoming the future of software development. But incidents like these outages suggest that the technology may not be ready for the level of trust and autonomy it’s being granted.

As regulators, industry leaders, and the public grapple with these questions, one thing is clear: the era of AI coding assistants is here to stay. The question is whether we can harness their potential without suffering the consequences of their current limitations.

More on AI: Look Out, OpenAI: Perplexity Admits AI Adverts Were a Mistake, Is Now Getting Rid of Them

Tags & Viral Phrases:
Amazon AI disaster, Kiro coding agent fails, internet outage caused by AI, AWS Christmas meltdown, AI autonomous coding gone wrong, tech companies’ AI addiction, Amazon blames users not AI, AI hallucinations in production, vibe coding catastrophe, Amazon’s AI mandate backfires, AI coding assistant reliability crisis, Amazon Web Services fails again, AI making decisions it shouldn’t, tech industry AI overhype, Amazon December 2024 outage, AI tools causing internet problems, Amazon’s AI blunder, Kiro autonomous agent disaster, Amazon AI coding tool controversy, AI coding tools unreliable, Amazon AI outage scandal, tech companies rushing AI implementation, Amazon AI tool permissions error, AI coding assistants producing buggy code, Amazon AI development crisis

,