By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.
On with Kara SwisherThe terrifying AI cyber threat no one is ready for
Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.
At a glance
WHAT IT’S REALLY ABOUT
AI is lowering the barrier to catastrophic cyberattacks at scale
- Nicole Perlroth argues that AI models like Anthropic’s Mythos can discover and chain software vulnerabilities, pushing advanced cyber exploitation toward broad availability.
- She details how North Korean operatives infiltrate U.S. companies via fake remote IT jobs, identity abuse, and U.S.-based laptop farms, often initially focused on steady income rather than immediate sabotage.
- She describes North Korea’s evolution from destructive attacks like Sony to sophisticated bank and crypto heists, framing cybercrime as a major source of regime funding.
- She says China has advanced from loud phishing to stealthy state operations that leverage zero-days and target security devices, telecom, and critical infrastructure such as water and power.
- She warns that U.S. election security and disinformation defenses are being weakened by agency dismantling, raising risks ahead of the midterms and reducing deterrence despite ongoing FBI efforts.
IDEAS WORTH REMEMBERING
5 ideasAI models are starting to automate high-end software exploitation.
She describes Mythos as a model built to write code that can also find and chain vulnerabilities across major operating systems and browsers, enabling actions like pulling photos from a locked phone or gathering banking activity via a malicious website.
The attacker pool is widening from a few elite actors to many.
Perlroth says capabilities once limited to top nation-state teams can now be used by lower-tier states and cybercriminals, increasing the number of actors who can break in “essentially anywhere they want.”
North Korea is scaling a remote-work infiltration playbook using identity fraud and laptop farms.
She says North Korean IT workers have been applying under stolen or borrowed identities, using U.S.-based “laptop farms” where an American hosts a corporate laptop while a North Korean remotes in.
The infiltration campaign is industrialized and high-volume, not isolated.
Perlroth reports that after gaining access to a Discord channel tied to one cell, a security firm observed that 22 people in that cell applied to 160,000 U.S. jobs in three months, revealing the operation’s scale.
Even ‘paycheck-only’ intrusions can evolve into theft and extortion once access becomes harder to sustain.
She says these hires often initially “don’t do anything” malicious beyond collecting paychecks, but as companies detect and fire them, they increasingly need to monetize access via extortion or data theft.
WORDS WORTH SAVING
5 quotesWhich means it can find vulnerabilities in every major operating system, every browser, and chain them together into an exploit so it can do things like...
— Nicole Perlroth
The problem is now these models can do what only previously the tier one elite nation-state intelligence or their contractors could do.
— Nicole Perlroth
But they have essentially become the masters of the blockchain, we just don't know it yet.
— Nicole Perlroth
The scale of this is insane.
— Nicole Perlroth
You know, we are locked in this very dangerous game of chicken.
— Nicole Perlroth
SPEAKERS
Nicole Perlroth
guestCybersecurity journalist and author of "This Is How They Tell Me the World Ends," formerly at The New York Times and an advisor to CISA.
Kara Swisher
hostJournalist and podcast host of "On with Kara Swisher."
High quality AI-generated summary created from speaker-labeled transcript.
