One Hacker, 72 Hours, and an Extorted AWS Environment

What Happened? 

Incident response firm Sygnia published research this week on a financially motivated attacker who used agentic AI workflows to speed up victim reconnaissance, attack tool development, command structuring, and adaptation to the target environment.

The attacker started with an AWS access key stolen through a weakness in an internet-facing application. From there, they pushed that key through four repeatable workflows built to grab as much data and access as possible. Every time new access turned up, it went back through the same workflows. Those workflows covered systematic secrets theft, backdoor creation, and data exfiltration.

Sygnia was clear that no single misconfiguration got exploited. The intrusion chained together weaknesses across application services, AWS resources, source code repositories, CI/CD pipelines, runtime components, and data stores. Credential discovery, secrets harvesting, cloud enumeration, deployment pipeline abuse, runtime modification, and database access all happened in parallel.

Then came the pressure campaign. The attacker mostly picked reversible moves to prove they could hurt the business without burning it down. They denied access to S3 buckets, scaled ECS services and containers down to zero capacity, wrote ACL rules to block network access, and purged SQS queues. Sygnia read it as a showcase of force, a signal that things could get far worse. The victim, an unnamed global enterprise, paid.

The Impact 

Here is the part that should worry security leaders. Researchers judged this workload as something that normally takes weeks. One person compressed it into 72 hours by letting AI handle reconnaissance, script writing, and on the fly decisions about a cloud environment they had never seen.

That collapses the old assumption that scale requires a team. A solo operator now carries the throughput of an organized group, which means the pool of people capable of hitting a large enterprise just got much bigger.

Avi Dayan, vice president of incident response at Sygnia, told Dark Reading that from a tactical view it does not matter whether AI wrote the command. Operationally, it changes everything. Mean time to detect and mean time to remediate have to contract sharply. If a tool can break out or exfiltrate data in under a minute, a team relying on humans triaging SIEM alerts will always lose.

Most security operations centers are still built around human review queues. That model was already strained. Against an attacker moving at script speed across nineteen different tasks at once, it does not hold.

How to Avoid This 

Sygnia’s guidance is unglamorous, which is usually a sign it works.

• Keep full visibility across assets and identities, including machine identities and service accounts. 
• Tighten identity controls so a single leaked key cannot walk sideways into your pipelines. 
• Lock down cloud and development environments, especially CI/CD and source repositories, since that is where secrets pile up quietly.
• Layer your controls so no one weakness opens the whole chain. Automate detection and response instead of routing every alert through a person at 3am.
• And write containment procedures before you need them. Delays cost more when the attacker never sleeps.