
Nov 20, 2026 · 53 min
Listen from 5:30
Listen at 5:30
AI compresses cyberattacks faster than defenses can adapt
E458: Sanaz Yashar: AI Agents Gone Rogue, the Next Cybersecurity Crisis, and the AI Arms Race
As AI lowers the time and expertise required to exploit vulnerabilities, organizations must decide how much autonomous action they can safely trust before a cascading incident.
- 1AI is shortening the path from vulnerability discovery to coordinated attacks on critical infrastructure.
- 2Defenders need attack-path visibility, stronger guardrails, and proactive prevention rather than isolated patching.
- 3Business costs, workplace habits, and limited training are slowing adoption of AI-assisted security defenses.
Don't miss
Yashar describes agentic remediation in practice, including an energy company that autonomously remediates more than 60% of its attack chains.
The brief
Sanaz Yashar, drawing on Unit 8200, FireEye, and Zafran, argues that AI is compressing the time from finding a vulnerability to weaponizing it.
The central mismatch is trust: AI can expose enormous numbers of vulnerabilities, but defenders remain reluctant to let it act autonomously in live environments.
Yashar shifts the focus from patching isolated flaws to mapping attack chains, finding critical junctions, and validating controls before an intrusion begins.
She predicts cybersecurity will worsen before it improves as AI lowers the barrier for inexperienced threat actors and weakens informal constraints on cyber conflict.
The practical answer is agentic remediation with guardrails and kill switches; Yashar says one energy company already remediates more than 60% of its attack chains autonomously.
Her warning is economic as well as technical: governments, insurers, and businesses may need to share the burden before a cascading infrastructure failure changes the calculus.
What was said on this episode
34 statements · 12 positive · 21 negative · 1 neutral
- Sanaz Yasharon AI controlPositive0:00
The nation controlling AI will control the world.
“The nation would control AI, will control the world.”
Listen at 0:00
AI can be used against societies lacking policy, guardrails, and control.
“It can be used against you if you are not having the right policy and the guardrails and if you don't control it.”
Listen at 0:12
Cybersecurity conditions will worsen before improving.
“I think it's going to get much worse before it gets better.”
Listen at 0:34
- Sanaz Yasharon AI-enabled cyberattacksNegative0:37
AI-enabled threat actors and agents will cause major cybersecurity incidents.
“We are going to have big incidents in cybersecurity caused by threat actors weaponized with AI or by AI agents.”
Listen at 0:37
- Sanaz Yasharon AI vulnerability scannersNegative2:34
AI scanners are finding zero-day vulnerabilities daily.
“Today, AI scanners finding zero days every day.”
Listen at 2:34
- Sanaz Yasharon AI economyNegative4:01
The AI economy cannot scale without adequate security development.
“AI economy cannot scale without developing security.”
Listen at 4:01
- Sanaz Yasharon AI-enabled cyberattack chainsNegative9:27
AI-enabled attackers can chain weaknesses across systems to infiltrate successfully.
“they actually use all the different weaknesses in all the different systems and make one chain that makes AI... successful to infiltrate and to do the mission.”
Listen at 9:27
- Sanaz Yasharon Patch-and-scan cybersecurityNegative10:14
Traditional patch-and-scan security is no longer sufficient.
“the time of patch and scan, all of it is completely over.”
Listen at 10:14
- Sanaz Yasharon AI attack-path simulationPositive10:30
Organizations should use AI to simulate attack paths and chains.
“we need to use AI to simulate what are those paths, those chains.”
Listen at 10:30
- Sanaz Yasharon AI cybersecurity alertsNegative11:50
Waiting for security alerts is too late against rapidly evolving AI threats.
“waiting for an alert to tell you something's wrong, it's too late.”
Listen at 11:50
- Sanaz Yasharon Attack-path preventionPositive12:03
Security programs should shift from vulnerability management to attack-path prevention.
“you should go from managing vulnerabilities and weaknesses to preventing those attack paths.”
Listen at 12:03
- Sanaz Yasharon AI geopolitical competitionNegative14:04
An intense AI war is currently occurring behind the scenes.
“there is a very, very intense AI war behind the scene.”
Listen at 14:04
- Sanaz Yasharon Frontier AI leadersNeutral14:24
Frontier AI leaders now bear greater responsibility for safeguards.
“those frontier AI leadership in the world has now more responsibility”
Listen at 14:24
- Sanaz Yasharon Agentic cybersecurity systemsNegative15:32
Developing a full agentic defense system is expensive for companies.
“It's expensive for each company to develop a full agentic system.”
Listen at 15:32
- Sanaz Yasharon Workforce AI trainingNegative17:11
The current workforce lacks training for efficient and safe AI use.
“Today, a workforce is not yet trained to use AI in an efficient and safe way.”
Listen at 17:11
- Sanaz Yasharon AI-driven job displacementNegative18:07
Some job roles will gradually disappear because of AI adoption.
“there are going to be roles that are going to be completely disappeared little by little”
Listen at 18:07
- Sanaz Yasharon AI-caused cybersecurity incidentsNegative21:12
Major AI-caused cybersecurity incidents will occur in the near future.
“We are going to have big incidents in cybersecurity caused. by threat actors weaponized with AI or by AI agents”
Listen at 21:12
- Sanaz Yasharon AI-enabled cyberattacksNegative23:03
AI is lowering the barrier to entry for cyberattacks.
“because of AI, we are seeing the bar is being lower.”
Listen at 23:03
- Sanaz Yasharon New cyber actorsNegative23:37
Many new cyber actors are emerging and growing rapidly.
“we are seeing like a lot of newcomers to the cyber world. And they are growing fast.”
Listen at 23:37
The PHP vulnerability affected approximately 70% of internet-facing websites.
“This PHP is in every website. It's in 70% of the websites that are internet-facing in the world.”
Listen at 25:26
- Sanaz Yasharon AI-enabled infrastructure attacksNegative26:18
A future incident exploiting similar weaknesses will have greater impact.
“next time it's going to be much more impactful.”
Listen at 26:18
- Sanaz Yasharon Preventive cybersecurityPositive31:29
Organizations should adopt more preemptive and preventive cybersecurity.
“you should be much more preemptive and preventive in advance.”
Listen at 31:29
- Sanaz Yasharon AI attack chainsPositive31:47
Organizations should implement guardrails and kill switches for attack chains.
“you have guardrails and you have kill switches for those chains.”
Listen at 31:47
- Sanaz Yasharon AI attack agentsNegative32:35
AI agents can assemble attack chains too quickly and unpredictably for humans to anticipate.
“they do it so fast. and so unpredictable that you cannot predict it by human eyes.”
Listen at 32:35
Existing tools can mitigate up to 90% of vulnerabilities in many environments.
“with the existing tools, they can take down up to 90% of vulnerabilities.”
Listen at 37:01
- Sanaz Yasharon Agentic remediation at an energy companyPositive38:56
An energy company autonomously remediates more than 60% of its attack chains.
“more than 60% of those attack chains are being remediated autonomously.”
Listen at 38:56
- Sanaz Yasharon Government AI security supportPositive41:12
Governments should initially supervise and subsidize AI security efforts.
“the governments should supervise and subsidize stuff.”
Listen at 41:12
- Sanaz Yasharon Catastrophic cyber eventNegative41:59
Waiting for a catastrophic cyber event would make recovery too late.
“when the Chernobyl will happen in cyber, it will be too late to fix it.”
Listen at 41:59
Security leaders now face a major new zero-day issue every day.
“Every day there is a big thing with a new zero day.”
Listen at 42:58
At least four companies were attacked simultaneously through the same vector.
“I see like at least four different companies being attacked by the same vector. At the same time”
Listen at 45:02
- Sanaz Yasharon AI-assisted cybersecurity trainingPositive47:11
AI agents can help junior workers become effective cybersecurity defenders.
“suddenly you see junior people in... that they do the first steps in security, use agents, and they become like cybersecurity defenders, like beast mode.”
Listen at 47:11
- Sanaz Yasharon Hiring practicesPositive48:29
Leaders should hire people stronger and more different than themselves.
“you should always, I think, hire people that they are better than you and they are much more different than you.”
Listen at 48:29
- Sanaz Yasharon Technology entrepreneurshipPositive51:57
Successful technology companies need focused, flexible, and agile people.
“the problem is those people that they know what they want to achieve and they are flexible and agile”
Listen at 51:57
- Sanaz Yasharon Startup foundersPositive52:49
Founders should listen less to investors and remain agile and mission-oriented.
“listen less to investors and be agile and mission oriented.”
Listen at 52:49
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.
Weekly email
State of the Pod, every Monday
The week's most-discussed names and stories from podcasts, with the moments behind them. Free, five minutes, unsubscribe in one click.