Palo Alto Networks CEO: "AI Found 5 Years of Bugs in 6 Weeks"

| Podcasts | June 08, 2026 | 207 Thousand views | 31:21

TL;DR

Palo Alto Networks CEO Nikesh Arora reveals that AI tool Mythos discovered 5 years of security vulnerabilities in just 6 weeks at low cost, declaring analytical SaaS 'dead' while warning that sophisticated AI cyber capabilities will be available to attackers within months, creating a critical race between defense and offense.

🔒 AI Security Breakthroughs 4 insights

Mythos found 5 years of bugs in 6 weeks

AI assessed Palo Alto's codebase and discovered vulnerabilities that would have taken 5-7 years to find manually, costing only low millions in tokens.

AI enables persistent vulnerability chaining

Using 'ultra mode' with persistent thinking, AI can daisy chain vulnerabilities to discover entirely new attack paths into systems.

Democratizing intelligence for consistency

AI allows 5,000 customer-facing employees to operate with 90% consistent quality rather than varying individual performance levels.

Attack capabilities approaching fast

Similar AI vulnerability-finding capabilities are 3 months away from being available in open-source models or Chinese alternatives.

💼 Enterprise Software Disruption 4 insights

Analytical SaaS is 'dead'

Companies no longer need third-party SaaS to analyze data when they can run LLMs directly against their own databases, reducing bills by 90%.

Infrastructure software undervalued

Storage and database companies (Snowflake, Databricks, MongoDB) will see 10x data growth as enterprises need massive memory to defend against AI attacks.

End of the UI era

Agents will replace traditional user interfaces, automatically processing Zoom calls, emails, and sales data without human data entry.

System of record reinvention

SaaS systems of work (Salesforce, etc.) must be re-engineered for agentic backends over the next 5 years as UI disappears.

⚠️ Cybersecurity Threat Landscape 3 insights

Defender race against time

Enterprises face a race to patch vulnerabilities before attackers use AI to exploit them, but most companies lack adequate memory and context of their systems.

Economic chaos over infrastructure

The greatest risk is economic disruption to small businesses (doctor offices, dentists) using legacy software, not just critical infrastructure like power plants.

Credential theft remains primary vector

89% of breaches currently occur from stolen credentials and simple password theft rather than sophisticated AI attacks.

🤖 AI Business Models & Strategy 3 insights

Models becoming utility layer

Intelligence will be commoditized with companies paying for specific 'IQ levels' (120 vs 250) rather than one-size-fits-all models.

Profit pools shift to applications

Value will accrue to application companies that arbitrage between models to solve business problems, not base model providers like OpenAI.

Enterprise data requirements multiply

Organizations must collect 10x more enterprise-wide data to establish baselines for 'good' vs 'bad' activity to defend against AI attackers.

Bottom Line

Organizations must immediately audit their codebases using AI vulnerability detection tools while consolidating data infrastructure to prepare for a future where AI agents both defend against and perpetrate cyber attacks within the next 3-6 months.

More from All-In Podcast

View all
GameStop CEO Ryan Cohen’s $56B Plan to Take Over eBay
1:03:03
All-In Podcast All-In Podcast

GameStop CEO Ryan Cohen’s $56B Plan to Take Over eBay

Ryan Cohen details his journey from founding Chewy to leading GameStop, explaining his contrarian investment philosophy, hands-on management approach prioritizing 'will over skill,' and the critical pivot away from his initial strategy to transform GameStop into a Chewy-like e-commerce business.

about 1 month ago · 10 points