Aikido promotes AI pentesting as a predictable-cost answer to faster release cycles and shrinking exploit windows
This Aikido vendor blog argues that AI-assisted coding has outpaced the annual or biannual manual pentest cycle, and presents AI pentesting with fixed per-test pricing as the fix. Citing its 2026 State of AI in Pentesting report, it says 76% of organizations ship significant production changes weekly or faster, while only 21% validate security on every release. It also cites Zero Day Clock data (via a16z) showing the share of exploited CVEs hit on or before disclosure day rising from 18.7% in 2018 to 53.6% in 2025 and 86.7% in 2026 so far. The post is mainly product marketing for Aikido Infinite, with no specific incident, vulnerability or threat actor.
Technical Details
- GitHub monthly commits reportedly grew from 1.4 billion (April 2026) to 2.9 billion (August 2026), against roughly 0.5 billion per month in 2023.
- AI coding agents still introduce flaws such as broken access control and business logic errors.
- Survey figures: 79% worry about vulnerabilities introduced between scheduled tests, 42% delay retests due to cost or scheduling, and only 40% retest promptly after a fix.
- Aikido claims its AI agents map the attack surface and validate findings through real exploitation. Its internal benchmarking says a full-scope test costs less than half of a human pentest, with results in hours versus a 2.6-week manual average.
- Aikido Infinite offers change-aware continuous testing that launches only on security-relevant changes, with a forecast, a prepaid credit wallet and spending limits.
Remediation
- Adopt risk-based testing: deep AI baselines plus incremental tests for frequently changing crown-jewel applications, and yearly baselines for low-risk ones.
- Retest fixes promptly rather than assuming they worked.
- Keep specialist human pentests for complex or high-assurance systems.
- Treat the vendor-supplied statistics and cost claims as unverified marketing.
Linha do tempo
-
Aikido promotes AI pentesting as a predictable-cost answer to faster release cycles and shrinking exploit windows
ler a análise desta fonte ocultar
This Aikido vendor blog argues that AI-assisted coding has outpaced the annual or biannual manual pentest cycle, and presents AI pentesting with fixed per-test pricing as the fix. Citing its 2026 State of AI in Pentesting report, it says 76% of organizations ship significant production changes weekly or faster, while only 21% validate security on every release. It also cites Zero Day Clock data (via a16z) showing the share of exploited CVEs hit on or before disclosure day rising from 18.7% in 2018 to 53.6% in 2025 and 86.7% in 2026 so far. The post is mainly product marketing for Aikido Infinite, with no specific incident, vulnerability or threat actor.
Technical Details
- GitHub monthly commits reportedly grew from 1.4 billion (April 2026) to 2.9 billion (August 2026), against roughly 0.5 billion per month in 2023.
- AI coding agents still introduce flaws such as broken access control and business logic errors.
- Survey figures: 79% worry about vulnerabilities introduced between scheduled tests, 42% delay retests due to cost or scheduling, and only 40% retest promptly after a fix.
- Aikido claims its AI agents map the attack surface and validate findings through real exploitation. Its internal benchmarking says a full-scope test costs less than half of a human pentest, with results in hours versus a 2.6-week manual average.
- Aikido Infinite offers change-aware continuous testing that launches only on security-relevant changes, with a forecast, a prepaid credit wallet and spending limits.
Remediation
- Adopt risk-based testing: deep AI baselines plus incremental tests for frequently changing crown-jewel applications, and yearly baselines for low-risk ones.
- Retest fixes promptly rather than assuming they worked.
- Keep specialist human pentests for complex or high-assurance systems.
- Treat the vendor-supplied statistics and cost claims as unverified marketing.