Alexi Carlier

Founder of Asymmetric Security and advocate for AI-driven defensive cybersecurity.

Claims

As speaker

Mentioned in

Key positions and views

The 'jagged frontier' of AI capabilities allows for the intentional acceleration of defensive applications over offensive ones by curating specific datasets and evaluation environments.

Digital forensics is the ideal domain to pursue differential acceleration because the human skill sets for forensics and offensive hacking have little overlap.

The primary cyber threats to organizations are bifurcated: high-frequency, low-sophistication attacks (phishing, unpatched vulnerabilities) and low-frequency, high-impact attacks from nation-states (IP theft).

The arrival of true AGI will be signaled not by passing abstract tests, but by its measurable impact on economic indicators like GDP as it begins to fully substitute human labor.

A major impediment to progress in defensive AI is the scarcity of deep forensic experts, whose small numbers create a bottleneck for evaluating and improving AI models.

Podcast consensus on Carlier

Points of consensus

▶Alexi Carlier is a strong proponent of using AI to differentially accelerate defensive capabilities in cybersecurity, particularly in the domain of digital forensics.Apr 2026

▶He characterizes the cyber threat landscape as bifurcated, with a large volume of attacks from unsophisticated, financially-motivated criminals and highly sophisticated, large-scale operations from nation-states like China focused on IP theft.Apr 2026

▶Carlier identifies significant bottlenecks to improving defensive AI, including a scarcity of deep forensic experts for evaluation, a lack of public data from security incidents, and cultural skepticism from incumbent cybersecurity firms.Apr 2026

▶He is described as "AGI-pilled," believing in the transformative potential of AGI while maintaining that current models are not yet AGI because they cannot reliably perform long-horizon tasks.Apr 2026

Points of debate

▶All claims originate from a single source, so there is no direct debate between sources. However, Carlier's own arguments present contrasting dynamics.

▶He emphasizes that the vast majority (around 80%) of cyber attacks are conducted by unsophisticated criminals, yet also highlights the massive, well-resourced threat from nation-states like China, creating a tension between the common threat and the most dangerous one.Apr 2026

▶Carlier's thesis for differential acceleration rests on the observation that in humans, the skill sets for forensics and hacking are distinct. It's an open question whether this distinction will hold as AI capabilities advance, potentially allowing a single advanced model to excel at both.Apr 2026

▶He is bullish on AI's potential to revolutionize defense but points out that the very industry he aims to transform is culturally skeptical and slow to adopt AI, posing a significant barrier to his vision.Apr 2026

Key themes

▶Differential Acceleration of Defensive AIApr 2026

Carlier's core thesis is that AI's development can be intentionally steered to benefit defenders more than attackers. He identifies digital forensics as an ideal domain for this, arguing that the skills for forensics do not easily translate to offensive hacking, creating an opportunity to build defensive moats by curating specialized datasets and evaluations.

This strategy of focusing on niche, defense-centric domains like forensics could offer a viable path for startups to outmaneuver larger incumbents and address the asymmetric nature of cybersecurity, suggesting a focus on specialized data and evaluation is a key competitive advantage.

▶The Bifurcated Cyber Threat Landscape

Carlier provides a nuanced view of cyber threats, separating them into two main categories. The first is the high-volume, low-sophistication threat from financially motivated criminals using common methods like phishing and unpatched vulnerabilities. The second is the high-sophistication, targeted threat from nation-states like China, which employs vast resources for objectives like intellectual property theft.

For analysts, this distinction is critical; solutions and risk models must account for both the common, everyday attacks that plague most organizations and the severe, targeted attacks that pose an existential threat to R&D-heavy industries.

▶Bottlenecks to AI in CybersecurityApr 2026

Carlier argues that the adoption of advanced AI in cybersecurity is being held back by several key factors. These include a severe talent shortage of deep forensic experts needed to evaluate AI systems, the reluctance of companies to share security log data, and the inherent skepticism of personnel at large incumbent firms towards the transformative potential of AGI.

Investors should recognize that the primary barriers to AI adoption in this sector are not just technical but also structural and cultural, and that companies which can solve the data access and human evaluation problems will have a significant market advantage.

▶Pragmatic AGI VisionApr 2026

Described as "AGI-pilled," Carlier is a firm believer in the future impact of Artificial General Intelligence. However, he maintains a grounded definition, stating current models are not AGI because they lack the ability to reliably string together tasks over long horizons. He proposes a concrete, economic benchmark for AGI's arrival: a measurable impact on GDP as it begins to substitute for human labor.

Carlier's focus on AGI's economic impact as the key indicator of its arrival provides a tangible, albeit lagging, metric for tracking its progress, shifting the conversation from abstract capabilities to real-world economic substitution.

Source episodes

Sentiment over time

Not enough data for timeline

Changes over time

2020-2021

Carlier was part of the early team at the Center for the Governance of AI think tank, indicating an early career focus on the strategic implications and governance of advanced AI.

Post-2021

Founded Asymmetric Security to apply his ideas about AI governance and differential acceleration to the domain of defensive cybersecurity.

Present

Actively promoting his company's AI platform, which claims to reduce email compromise investigation times from days to hours, and pursuing a go-to-market strategy through cybersecurity insurance providers.

Present

Articulates a vision for an "AGI-pilled" approach to cyber defense, arguing that the field is ripe for disruption due to incumbent skepticism and a scarcity of human forensic talent.

Suggested prompts

How viable is Alexi Carlier's strategy of 'differential acceleration' for defense, and what are the primary risks that could undermine it? &nearr;To what extent does Asymmetric Security's go-to-market strategy through insurance companies mitigate the data scarcity problem Carlier identifies? &nearr;How does Carlier's distinction between unsophisticated criminal hackers and sophisticated state actors influence his product development and threat modeling? &nearr;What evidence supports or refutes Carlier's claim that large cybersecurity incumbents are too skeptical of AGI to effectively innovate? &nearr;

Key concepts

Digital Forensics 4 ep AGI 4 ep Cyber Threat Landscape 4 ep Business of Cybersecurity 4 ep Nation-State Hacking 2 ep Differential Acceleration 2 ep AI Training Bottlenecks 2 ep Ransomware 1 ep

Notable quotes

“the implication here, I think, is that actually you can go out and pick various areas that you think it might be important to harden the world, like defensive cybersecurity... And you could very intentionally curate the datasets, the environments, the evals that you need to sort of like pull out the jagged frontier in this specific direction.”

Alexi Carlier · AGI-Pilled Cyber Defense: Automating Digital Forensics w/ Asymmetric Security Founder Alexis Carlier

“You have China, a big one, obviously the CCP. A big focus of theirs is IP theft from R&D heavy industries in the West. AI is almost definitely like a huge focus now.”

Alexi Carlier · AGI-Pilled Cyber Defense: Automating Digital Forensics w/ Asymmetric Security Founder Alexis Carlier

“Relatively recently, Jaguar had a big ransomware case. And if I recall correctly, they needed a $2 billion loan from the UK government after this happened to avoid going under. And their operations were down for a month, two months, something like that.”

Alexi Carlier · AGI-Pilled Cyber Defense: Automating Digital Forensics w/ Asymmetric Security Founder Alexis Carlier

“Current AI models are not yet AGI because while they are superhuman at some tasks, they cannot reliably string together tasks over long horizons.”

Alexi Carlier · AGI-Pilled Cyber Defense: Automating Digital Forensics w/ Asymmetric Security Founder Alexis Carlier

Report last updated: Apr 21, 2026

Get started free

Back to Entities Intelligence Report