Sam Altman: AI Security Is the Next Big Tech Frontier

The world’s most influential AI leader and OpenAI CEO Sam Altman has issued his clearest warning yet: AI security is about to become one of the most important—and fastest-growing—fields in technology. Speaking with Stanford professor and cryptography expert Dan Boneh, Altman said the world is entering a phase where AI safety concerns will “recast” themselves as urgent security challenges with global implications.

“AI security is probably a very, very undervalued field right now,” Altman said.
“Given how capable these models are getting, the security problems are going to get really big.”

His comments create a new industry news peg: the world’s most influential AI leader now openly signals that AI security will shape workforce demand, corporate risk, national infrastructure, and technology innovation for the next decade.

AI Security Becomes Critical Infrastructure

Altman told Boneh that as AI systems become more capable and more personalised, organisations must prepare for entirely new classes of vulnerabilities.

He identified one emerging threat with especially high stakes:
AI models that learn deeply personal information while also interacting with external services.

“People love how personalized these models are getting… and they love that [AI] can call things on the web,” Altman said. “But you don’t want someone to exfiltrate data from your personal model that knows everything about you.”

Humans naturally adjust what they say based on context. AI systems do not.
Altman warned that when people tell an AI about their health, finances or personal life, they may also ask that same system to transact online.

“You don’t want that e-commerce site to know all of your health issues,” he said. “This is a very interesting security problem to solve with 100% robustness.”

His message: AI personalisation + tool use = a new frontier of security risk.

And it will require thousands of new specialists.

AI Security Will Be One of the Fastest-Growing Career Sectors

Boneh asked whether AI security is a good field for developers, cybersecurity professionals, or students entering tech.

Altman didn’t hesitate:

“I think this is one of the best areas to go study.”

He compared the moment to the rise of internet security in the early 2000s, noting that cloud giants like Google and Amazon employ thousands of engineers dedicated to platform security.

AI companies, he said, will need the same scale—likely larger.

AI Will Soon Work as a “Superhuman” Security Analyst

Altman also predicted that AI will transform offensive and defensive cybersecurity operations.

Boneh pointed out that systems like ChatGPT already identify bugs and vulnerabilities. Altman confirmed that OpenAI is accelerating these capabilities:

“We should be a superhuman AI security analyst pretty soon.”

He expects AI systems to:

• audit code
• find vulnerabilities
• test software before deployment
• provide defence against large-scale cyber attacks
• and, inevitably, power advanced offensive tools

“It works both directions,” Altman said. “It’s going to be a big deal for cyber attacks.”

For organisations worldwide, this signals a shift: AI will write code, test code, break code, and secure code—all at scale.

Software Development Will Shift to “Talking to a Computer”

Altman used blunt language to describe the future of software engineering.

“It’ll mostly be like talking to a computer.”

He expects developers to describe the software they want in English or pseudocode, and AI systems will:

• generate code overnight
• build full applications
• write tests
• correct failures
• and maintain repositories through autonomous engineering agents

This transition could redefine the global developer workforce—boosting productivity while elevating AI fluency as the core career skill.

Computer Science Education Must Change—Fast

Altman said universities worldwide must rethink their curricula.

He values fundamentals like C++ and operating systems, but he believes current courses lag the real frontier:

“Intro to programming should probably change quite a lot.”

Today, training neural networks sits at the cutting edge.
In 10 years, he predicts even that may become a background skill:

“At some point you’ll just tell the AI to train a new neural network.”

He emphasised that the most valuable future skill will be:
knowing how to use AI to do remarkable things.

“Early Innings”: AI Architecture, Energy, and Hardware Still Wide Open

Altman rejected the idea that transformers or current deep learning architectures represent an end point.

“I don’t feel like we’re close to the end of the roadmap.”

He pointed to massive open research opportunities:

• new AI architectures beyond transformers
• 100× improvements in energy efficiency
• optical computing
• improved cooling technologies
• new training and inference algorithms

He described the field as “virgin territory.”

A Message to the World’s New Generation of Technologists

Altman delivered one of the strongest endorsements of an AI-focused career seen from any industry leader:

“This is probably the best time in my lifetime to be early in a computer science career.”

His advice: work with smart, optimistic people on problems that matter.

“You have the opportunity for this to be the most important work you ever touch in your life.”

And in a line that captures the spirit of the moment:

“AI is the main technology being developed… so why would you not be part of this?”

To explore every insight from this conversation, watch Sam Altman’s full interviewas part of the Stanford Advanced Cybersecurity Program.