“Cybersecurity Superintelligence – the missing piece of AGI’s puzzle”

– By Saket Modi
Co-Founder and CEO at SAFE

 

AI has been hailed by many visionaries as the greatest breakthrough of our time – some even say of all time. Whether or not that proves true, one thing is certain: the world will never be the same.

In this new era, AI unlocks unprecedented opportunities. It promises to elevate humanity – by creating abundance, accelerating the global economy, and pushing the frontiers of scientific discovery. It will free us from mundane, repetitive tasks and unleash a new age of creativity, curiosity, and exploration into the very nature of life itself.

But not everything AI brings will be progress.

AlphaGo’s legendary Move 37 – an unthinkable play beyond human comprehension—changed the game of Go forever. But what happens when a “Move 37” is made against humanity, this time by an AI?

The consequences could be far more devastating than a game lost – potentially darker than doomsday itself.

What Is Move 37?

15th March 2016 – in the AlphaGo vs. Lee Sedol match, Move 37 was a turning point in AI history. AlphaGo, playing as black, placed a stone on the 5th line from the edge – a highly unconventional position early in the game. This was Move 37. And why was it iconic?

1. Humans wouldn’t have played it: It defied traditional Go strategy; experts initially thought it was a mistake.
2. Creative, not brute-force: It showcased that AI could be creative, not just calculating.
3. Changed the outcome: It shocked Lee Sedol and shifted the momentum, leading AlphaGo to win that game.
4. Symbolic moment: It marked the moment AI surpassed even the most intuitive human mastery in a deeply complex game.
   

In essence, Move 37 wasn’t just a move – it was a message: AI doesn’t think like us, and it can still beat us.

What happens when AI makes a move against humanity – starting in its own “home turf,” the cyber world—that’s now inseparably intertwined with our physical reality?

The rise of AGI will demand a new kind of counterforce – an “Alter-AI” to keep it in check. As the new makers reshape our world, we’ll need new guardians to protect it – the maker always needs a checker. In this new era, CyberAGI must emerge as the balance.

OpenAI defines AGI as systems that surpass human intelligence and benefit all of humanity – autonomous agents that can do most things better than most people, safely. I agree – but that’s just one story of AGI.

If there’s one thing Marvel taught us: there’s never just one hero. There’s always another – sometimes the savior, sometimes the reckoning.

The Case for SuperSpeciality AGI

Take the smartest human of the last 100 years – say, Albert Einstein. If we use AGI’s definition – someone who can do most things better than most humans – Einstein wouldn’t qualify. We all know that’s not true.

But where he did surpass almost every human was in Physics and Mathematics. His neural network—biological, not artificial—was trained on the leading research of his time. He didn’t just understand it; he reasoned through it to uncover truths the world had never seen before.

In that sense, Einstein wasn’t AGI. He was PhysicsAGI.

Meta’s Chief AI Scientist, Yann LeCun, has argued that AGI doesn’t truly exist—because even human intelligence is highly specialized.

We agree. In our view, there will be general-purpose AGI trained on vast, publicly available data—led by players like OpenAI, Meta, and Google.

But the real breakthrough will come from domain-specific AGI—deeply specialized systems trained on proprietary, contextual data. These systems won’t just mimic intelligence; they’ll reason with expert-level precision within their fields.

The way in our view, 

• HealthAGI aims to solve the ultimate challenge of the physical world: a disease-free future
• PhysicsAGI will help us unlock the universe’s deepest mysteries—through pure reasoning.
  

And then there’s CyberAGI—built to solve the defining problem of the digital age: making companies and individuals unhackable.

So What Is CyberAGI?

Cyber Artificial General Intelligence (CyberAGI) refers to a system of intelligence capable of autonomously predicting, detecting, and remediating any threat to the confidentiality, integrity, or availability of digital systems and individuals.

For businesses, CyberAGI spans several domains—aligned with the NIST Cybersecurity Framework—and requires deep contextual understanding across:

1. Business context & asset inventory
2. Third-party risk
3. Threat intelligence
4. Identification measures
5. Protection measures
6. Detection findings
7. Response strategies
8. Recovery mechanisms
9. Governance, risk & compliance

Key Capabilities of CyberAGI:

1. Generalization Ability: The power to transfer knowledge across domains, enabling rapid adaptation to new and unforeseen cyber threats.
2. Common Sense Reasoning: A foundational understanding of real-world facts, relationships, and digital behaviors, allowing CyberAGI to make contextually sound decisions.
3. The development of CyberAGI is inherently interdisciplinary—drawing from risk analysis, detection and response, backup and recovery, threat intelligence, exploit research, and more. Progress in each of these areas is accelerating our path toward truly autonomous cyber defense.
   

The pursuit of CyberAGI demands deep interdisciplinary collaboration—across risk analysis, detection and response, backup and recovery, threat intelligence, exploit research, and more. Advancements in each of these domains are continuously shaping both the vision and the reality of CyberAGI.

CyberAGI – All or nothing?

Self driving has 6 Levels of Autonomous Driving; Google defines 6 Levels of AGI

We define 6 Levels of CyberAGI

Level	AGI	CyberAGI
Level 0 : No AI	Human in Loop Computing like Cruise Control	Human in Loop Cybersecurity with Basic Workflows
Level 1 : Emerging	Better than an Unskilled Human	Better than an Unskilled Human
Level 2 : Competent	Better than 50th Percentile of Skilled Adults	Better than 50th Percentile of Skilled Cybersecurity Analysts
Level 3 : Expert	Better than 90th Percentile of Skilled Adults	Better than 90th Percentile of Skilled Cybersecurity Analysts
Level 4 : Virtuoso	Better than 99th Percentile of Skilled Adults	Better than 99th Percentile of Skilled Cybersecurity Analysts
Level 5 : SuperHuman	Outperforms 100% of Humans in All Tasks	Outperforms 100% of CyberAnalysts in All Tasks

AGI researchers agree we are at Level 1 for AGI, and we think it’s the same level we are at for CyberAGI too. 

CyberAGI will predict and remediate hacks before they occur – even the kinds that the world has never seen – the zero days.

At SAFE, we’ve been quietly laying the foundation for CyberAGI—one building block at a time: CRQ, CTEM, TPRM, and more on the horizon. Today, we operate as a layer of intelligence and autonomous action across the cybersecurity stack for hundreds of Fortune 2000 companies worldwide, powering truly risk-based decision-making.

By not being a detection vendor, we remain uniquely neutral—laser-focused on helping our customers manage cyber risk with clarity, precision, and speed.

We believe the conditions are just right—a perfect storm that positions us to be the first company to build true cybersecurity superintelligence

And so…we take the boldest step in our journey…
Today, we elevate the mission of SAFE – We Will Build CyberAGI.

Read our press release:

SAFE Raises $70 Million Series C to Build CyberAGI; Unveils World’s First Fully Autonomous CTEM Solution