# How a security exec stopped cyberattack while walking his dog

Corma's defense AI detects and blocks threats in under 10 minutes

- **Source:** [www.theregister.com - Articles](https://brifing.si/source/2c34ea50-d341-4a10-925d-12c012a9ad10)
- **Published:** 2026-08-16
- **Topic:** [Data & AI](https://brifing.si/c/data-ai)
- **Original article:** <https://www.theregister.com/security/2026/08/16/stopping-a-cyberattack-while-walking-your-dog-defensive-ai-security-ceo-says-its-not-ruff-to-do/5288126>
- **On Brifing:** https://brifing.si/s/d3c65bc8-80ad-4707-8851-2982ab2212f6

## CEO's real-time intervention during dog walk

Corma CEO Alon Pluda recounts how a client stopped a cyberattack via smartwatch alert while walking his dog. The AI agent detected a live intrusion, sought approval, and secured the network. This incident showcases defense AI's critical role in real-world scenarios. Corma's technology neutralized the threat in under 10 minutes, demonstrating rapid response capabilities that complement human security teams.

## $60M defense 'ring' funding round

Corma secured $60M in seed funding led by Sequoia Capital, with participation from Khosla Ventures and Coatue. The company references 'Lord of the Rings' with its mission to build defensive superintelligence. This investment will accelerate partnerships with Fortune 100 companies. Pluda notes current models show exponential improvement in offensive capabilities while lagging in defense, creating what Corma calls the 'defense gap.'

## Why models underperform in defense

Corma's research reveals top models like GPT-5.5 achieve 85% success in attack simulations but only 19% in defense. This stems from training data over-indexing on code scanning rather than analyzing logs, configurations, and disk states - core to defensive work. The company develops specialized agents to bridge this gap. Pluda explains: 'Defensive security deals with structured machine data that models read less reliably.'

## Cross-industry AI security deployment

Fortune 500 companies in healthcare, finance, and energy have deployed Corma's AI workforce. Early data shows 94% faster threat response and 15x expanded security coverage. The agents detect multi-stage attack campaigns human teams miss. 'With proper optimization, we can surpass human intelligence's limitations,' says Pluda. The system functions as customizable digital security staff, assigned to specific defense tasks.

## Attack-defense simulation insights

In 241 test scenarios across 4 models, attackers established persistent backdoors 85% of the time. The same models detected only 19% of attacks when defending. This imbalance reflects inherent biases in foundational model training. Corma addresses this with defense-specific data pipelines. The company's architecture processes machine-generated data more effectively than general-purpose models.

## Three advantages of agentic defense

1) Real-time monitoring of multi-step attacks 2) Continuous analysis of logs and config changes 3) Scanning data points inaccessible to human teams. Corma deploys agents as 'virtual security staff' that organizations can task specifically. The system expands coverage across security functions while reducing operational costs. Early adopters report uncovering sophisticated attack campaigns previously undetectable.

## Why defensive security is more complex

Pluda contrasts offensive security's clear objectives ('break this system') with defense's open-ended 'needle-in-haystack' nature. Models struggle with unstructured machine data that dominates defensive work. Corma's solution involves purpose-built agent architectures optimized for parsing audit trails and disk states. These agents identify anomalies faster than human analysts by processing machine-native data formats.

## Balancing the cyber warfare equation

Corma positions AI-augmented defense as transformative for cybersecurity. Data shows AI agents dramatically reduce response times compared to human teams. 'We must ensure defenders win this intelligence-versus-intelligence war,' states Pluda. The industry anticipates proliferation of defense-optimized language models and simulation platforms to counter increasingly sophisticated AI-powered threats.

## Sources covering this story

- https://www.theregister.com/security/2026/08/16/stopping-a-cyberattack-while-walking-your-dog-defensive-ai-security-ceo-says-its-not-ruff-to-do/5288126 — 

---

_Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._