Latest / Ctrl AI Profit / Ep. 119 | AI Just Hacked a Company by Itself
Transcript
- 0:20SPEAKER_01: An AI just hacked a company by itself, not with a person directing it, not with a script someone wrote.
- 0:26SPEAKER_01: The AI made the decisions, it broke in, stole credentials, moved through the network, and walked out with an entire database in under an hour.
- 0:35SPEAKER_00: Security researchers at Sysdig published the forensic report last week, and it is being called the first confirmed fully autonomous LLM agent cyber attack ever recorded in the wild.
- 0:46SPEAKER_00: Not a lab test, not a red team exercise, a real intrusion against a real target driven end-to-end by an AI.
- 0:54SPEAKER_01: Let me make sure people understand what that means.
- 0:57SPEAKER_01: When you hear about a hacking attack, you usually picture a person sitting at a keyboard, typing commands, making decisions.
- 1:03SPEAKER_01: This was not that.
- 1:04SPEAKER_01: An AI agent, a large language model, was running the entire operation.
- 1:09SPEAKER_01: It decided what to look for, it decided where to go next, it adapted when something did not work, and it finished the job in about 60 minutes.
- 1:17SPEAKER_00: The attack chain is worth understanding because it shows how capable these agents already are.
- 1:22SPEAKER_00: Step one, the AI exploited a public-facing Merimo notebook using a known vulnerability.
- 1:28SPEAKER_00: That is the entry point.
- 1:30SPEAKER_00: Once inside, the agent started looking around.
- 1:33SPEAKER_00: It found cloud credentials stored on the compromised machine.
- 1:36SPEAKER_01: So the AI is already inside, and it immediately starts rummaging through the filing cabinets.
- 1:42SPEAKER_01: That is what credential discovery looks like.
- 1:45SPEAKER_01: It found two sets of cloud credentials, basically keys to other doors in the building.
- 1:50SPEAKER_00: Step two, using those credentials, the AI pivoted to cloud infrastructure.
- 1:56SPEAKER_00: It accessed AWS Secrets Manager and pulled down an SSH private key.
- 2:00SPEAKER_00: That is the key to yet another door, this one leading to a Bastion server, which is like a gateway to the most sensitive part of the network.
- 2:08SPEAKER_00: And step three?
- 2:11SPEAKER_00: The AI used that SSH key to open eight separate sessions against the Bastion server.
- 2:16SPEAKER_00: From there, it discovered an internal Postgres QL database and exfiltrated the entire thing.
- 2:22SPEAKER_00: Schema and contents.
- 2:24SPEAKER_00: The Bastion phase exfiltration completed in under two minutes.
- 2:28SPEAKER_00: Full chain, about 60 minutes.
- 2:31SPEAKER_01: Under two minutes to dump a database.
- 2:34SPEAKER_01: Think about that from the small business side.
- 2:36SPEAKER_01: You have a small medical practice, a law firm, a financial advisory, your patient records, client files, financial data gone in less time than it takes to brew a pot of coffee.
- 2:47SPEAKER_00: And here is what makes this fundamentally different from a traditional attack.
- 2:51SPEAKER_00: When a human hacker is inside your network, there are tells.
- 2:55SPEAKER_00: They hesitate, they research, they make mistakes, an LLM agent does not hesitate.
- 3:02SPEAKER_00: It runs at machine speed, adapts in real time based on what it finds, and never gets tired or distracted.
- 3:09SPEAKER_01: Sysdig identified several things that proved this was AI-driven, not human-driven.
- 3:14SPEAKER_01: Can you walk through those?
- 3:16SPEAKER_00: Yes, and this is the forensic part that should worry every business owner.
- 3:20SPEAKER_00: First, machine-shaped command output.
- 3:23SPEAKER_00: The commands were generated in response to prior output, not prescripted.
- 3:28SPEAKER_00: That means the agent was reading the results of each step and deciding what to do next.
- 3:33SPEAKER_00: Second, planning comments embedded in the session.
- 3:36SPEAKER_00: These look like internal agent thoughts, step annotations that are typical of how LLM tools scaffold their reasoning.
- 3:43SPEAKER_00: Third, dynamic branching.
- 3:45SPEAKER_00: The agent changed its approach based on whether commands succeeded or failed.
- 3:49SPEAKER_00: A static script does not do that.
- 3:52SPEAKER_01: So it was thinking, not like a human thinks, but making decisions based on what it observed, adjusting its plan, trying different paths.
- 4:01SPEAKER_01: That is the definition of autonomy.
- 4:05SPEAKER_00: There is one more detail that is critical.
- 4:07SPEAKER_00: The agent used Cloudflare workers as a distributed egress pool.
- 4:12SPEAKER_00: That means it was routing its traffic through multiple exit points to hide where the attack was actually coming from.
- 4:17SPEAKER_00: That is not basic script behavior.
- 4:20SPEAKER_00: That is operational tradecraft.
- 4:22SPEAKER_01: Okay, so now the question every small business owner should be asking.
- 4:26SPEAKER_01: If an AI can do this to a company with real cloud infrastructure, what can it do to my business?
- 4:34SPEAKER_01: Because most small businesses do not have a security operations center.
- 4:39SPEAKER_01: Most do not have someone watching the network at two in the morning.
- 4:43SPEAKER_00: That is exactly the point security professionals are making.
- 4:46SPEAKER_00: CrowdStrike's latest threat report shows the average breakout time for human-driven intrusions is 29 minutes.
- 4:53SPEAKER_00: This AI-driven attack completed its full chain in about 60 minutes, but that includes reconnaissance and decision making that a human would have done much more slowly.
- 5:01SPEAKER_00: The AI compressed the decision loop.
- 5:03SPEAKER_01: And the AI will only get faster.
- 5:06SPEAKER_01: This is the slowest, least capable version of this kind of attack that will ever exist.
- 5:11SPEAKER_01: Every month from now, the models get better, the agents get faster, and the tools get more accessible.
- 5:18SPEAKER_00: That accessibility point is crucial.
- 5:21SPEAKER_00: The barrier to launching an attack like this is dropping fast.
- 5:24SPEAKER_00: A year ago, you needed significant technical skill to execute a multi-stage intrusion.
- 5:30SPEAKER_00: Today, an LLM agent can do the heavy lifting.
- 5:33SPEAKER_00: Tomorrow, the agent itself could become a service.
- 5:36SPEAKER_00: Rent an AI hacker for an hour.
- 5:39SPEAKER_01: AI hacking as a service.
- 5:41SPEAKER_01: That is where this goes.
- 5:43SPEAKER_01: And when the cost of launching a sophisticated attack drops to near zero, the volume of attacks goes through the roof.
- 5:49SPEAKER_01: Small businesses become the easiest targets because they have the least defenses.
- 5:54SPEAKER_00: The security community is already shifting its language around this.
- 5:58SPEAKER_00: Experts are recommending that AI agents be treated as first-class identities, registered, authenticated, monitored, with guardrails and recovery plans.
- 6:06SPEAKER_00: That applies whether the agent is yours or an attacker's.
- 6:10SPEAKER_00: What does that mean practically?
- 6:11SPEAKER_01: Treating an AI agent like an identity?
- 6:14SPEAKER_00: It means every AI tool or agent that has access to your systems should have its own credentials, its own permissions, and its own audit trail.
- 6:23SPEAKER_00: Just like you would not let a new employee access everything on day one without oversight, you should not give an AI agent blanket access to your infrastructure.
- 6:32SPEAKER_01: That is a framework most small businesses have never even thought about.
- 6:36SPEAKER_01: They are still thinking about locking the front door.
- 6:38SPEAKER_01: They need to start thinking about who or what is already inside the building.
- 6:44SPEAKER_00: There is a specific lesson in how this attack started.
- 6:47SPEAKER_00: The entry point was an Internet exposed Marimo notebook with an unpatched vulnerability.
- 6:52SPEAKER_00: CVE 202630-9987 was already known.
- 6:57SPEAKER_00: CISA had added it to the known exploited vulnerabilities catalog.
- 7:01SPEAKER_00: The fix was available, but the target had not applied it.
- 7:04SPEAKER_01: Patch your stuff?
- 7:06SPEAKER_01: I know that sounds basic, but it is the single most important thing a small business can do right now.
- 7:11SPEAKER_01: Every unpatched system is an open door, and now there is an AI that can walk through it faster than any human ever could.
- 7:18SPEAKER_00: The second lesson is about credentials.
- 7:20SPEAKER_00: The AI found cloud credentials on the compromised machine and immediately used them to escalate the attack.
- 7:26SPEAKER_00: If those credentials had not been there, or if they had been rotated or scoped to minimal permissions, the attack chain would have broken at step two.
- 7:35SPEAKER_01: So the playbook for small businesses is actually straightforward, even if the threat is sophisticated.
- 7:41SPEAKER_01: One, patch everything, always.
- 7:44SPEAKER_01: Two, do not store credentials where an attacker or an AI can find them.
- 7:50SPEAKER_01: Three, limit permissions so that even if someone or something gets in, they cannot reach the crown jewels.
- 7:57SPEAKER_01: Four, monitor your systems so you know when something is moving through them.
- 8:01SPEAKER_00: And five, start thinking about AI agents as part of your security model, not just as tools your employees use.
- 8:08SPEAKER_00: If agents can attack you, agents should also be part of your defense.
- 8:12SPEAKER_00: AI-powered security monitoring is no longer optional for businesses that handle sensitive data.
- 8:17SPEAKER_01: Here is the thing that keeps me up at night.
- 8:20SPEAKER_01: We talk about AI as this incredible productivity tool, and it is, but every technology cuts both ways.
- 8:27SPEAKER_01: The same capabilities that let you automate your workflow let someone else automate an attack on your business.
- 8:33SPEAKER_01: The difference is you have to sleep.
- 8:36SPEAKER_01: The AI does not.
- 8:38SPEAKER_00: That asymmetry is the core challenge.
- 8:41SPEAKER_00: An AI agent can probe your defenses 24 hours a day, seven days a week, adapting in real time.
- 8:49SPEAKER_00: No human security team, especially not one at a small business, can match that pace without their own AI tools.
- 8:56SPEAKER_01: So the message is not panic, the message is evolve.
- 9:00SPEAKER_01: If you are not using AI to help protect your business, you are fighting a drone with a flashlight.
- 9:06SPEAKER_01: The technology exists, the attacks are real.
- 9:08SPEAKER_01: The question is whether you adopt the defensive side before the offensive side shows up at your door.
- 9:13SPEAKER_00: We will keep tracking this story as it develops.
- 9:16SPEAKER_00: The security landscape just fundamentally changed.
- 9:19SPEAKER_00: And every business, regardless of size, needs to understand that.
- 9:23SPEAKER_01: Check your patches, check your credentials, and if you do not have some form of AI powered monitoring on your network, make that your next investment, not next quarter.
- 9:35SPEAKER_01: Now