How AI could change the economics of cyber security

Key takeaways AI could change the unit economics of software: inference and specialised compute raise costs for vendors without hyperscale infrastructure, while potentially empowering customers to build more in-house. Early cyber security disruption will hit particular sub-sectors: static analysis and vulnerability discovery appear most exposed; endpoint, identity and recovery are more platform- and process-dependent, in our view. AI expands the attack surface, raising aggregate cyber security demand: convincing phishing, automated intrusion attempts and agent deployments will potentially increase demand for strong identity, monitoring and incident response.

For years, many software businesses benefited from a powerful business model: low cost of goods sold, high gross margins and recurring revenue. AI complicates that structure.

Running AI features costs real money (compute, model calls and engineering to manage them). Hyperscalers can amortise infrastructure and drive down cost per token; many specialist vendors can’t. At the same time, customers may choose to buy access to a large language model (LLM) and build bespoke tools rather than pay for packaged software.

In cyber security, this dynamic is already pushing ‘security as a feature’ deeper into the software development lifecycle. 

In the medium term, AI-native security tools are likely to become a standard layer embedded in how software is built, rather than an overnight replacement for existing security platforms.

Cost and access: why pricing signals strategic intent

Pricing shared for Mythos has been positioned at around $25 per million input tokens and $125 per million output tokens once an initial $100 million credit pool is exhausted (a rough rule of thumb is that one token is about three-quarters of a word). 

The numbers matter less than what they imply: high-performance security reasoning depends on compute, and compute turns ‘software margins’ into something closer to an infrastructure business.

Why cyber security is a headline theme right now

While the economics of cyber security software are shifting, in aggregate AI is likely to support the strong demand outlook for the theme, in our view.

• AI-powered social engineering is scaling: phishing has become more convincing and more automated. It’s estimated that the number of active ransomware groups has risen by 49% in 2026 compared with a year ago.[1] 
• Policy attention is rising: the UK has highlighted the economic impact of cyber breaches and announced a £210m Government Cyber Action Plan.[2]
• Geopolitical volatility increases risk: state-sponsored activity and supply-chain style techniques continue to pressure critical networks.
• Market growth remains strong: the global AI-in-cybersecurity market is projected to grow at a 21% compound annual growth rate between 2026 and 2034.[3]

In summary, AI is not ‘ending’ cybersecurity – it is changing what gets automated, shifting spend toward identity, monitoring and response, and forcing every organisation to build stronger remediation muscles. 

We believe the winners will pair AI-driven discovery with disciplined operations that can actually close the loop.

 

*For illustrative purposes only. Reference to a particular security is on a historic basis and does not mean that the security is currently held or will be held within an L&G portfolio. The above information does not constitute a recommendation to buy or sell any security.

[1] Source: IBM
[2] Source: Government Cyber Action Plan - GOV.UK
[3] Source: https://www.fortunebusinessinsights.com