EN PT-BR
← Voltar ao feed
low

Aikido launches AI-agent external network pentesting product, citing PHP2Shell exploitation speed

1 fonte(s) · 1 atualização(ões) · publicado 28/10/2026 07:00 · coletado há 4 dias

Aikido Security announced an external network pentesting product. AI agents map a customer's internet-facing assets from a single domain and attack them the way an outside adversary would. The vendor uses the PHP2Shell critical PHP flaws as its motivating example. It says Cloudflare saw thousands of exploit attempts in the 10 hours before disclosure, and millions by the next day. It also cites Zero Day Clock data charted by a16z: 86.7% of exploited vulnerabilities in 2026 were hit on or before disclosure day, versus 18.7% in 2018. This is a product announcement with no new incident, but it carries some vendor-reported exploitation statistics.

Detalhes técnicos

  • Agents enumerate subdomains and internet-facing systems (websites, APIs, VPN gateways, mail servers, self-hosted tools such as GitLab). They then attack each host using known CVEs, misconfigurations and exposed secrets.
  • Findings are validated and reproduced by other agents before reporting, and a supervisor agent blocks potentially disruptive actions.
  • Example attack chain: an exposed .git folder leaks repository history, which contains SSH credentials. These work on another host, where Azure credentials are found, giving admin access to the cloud tenant.
  • The article also references CISA's 48-hour disconnect directive for Ivanti VPN appliances in 2024.
  • Aikido's 2026 State of AI in Pentesting report (400 leaders surveyed) says 48% find pentest findings outdated on arrival.

Remediação

  • Test the external perimeter continuously rather than once or twice a year, and retest after fixes.
  • Inventory internet-facing assets and subdomains, and remove exposed .git folders, secrets and misconfigurations.
  • Patch quickly or take exposed systems offline when critical flaws are disclosed.

Linha do tempo

  1. aikido.dev ↗ · pub. 28/10/2026 07:00 · coletado 28/10 08:00 primeiro registro

    Aikido launches AI-agent external network pentesting product, citing PHP2Shell exploitation speed

    ler a análise desta fonte

    Aikido Security announced an external network pentesting product. AI agents map a customer's internet-facing assets from a single domain and attack them the way an outside adversary would. The vendor uses the PHP2Shell critical PHP flaws as its motivating example. It says Cloudflare saw thousands of exploit attempts in the 10 hours before disclosure, and millions by the next day. It also cites Zero Day Clock data charted by a16z: 86.7% of exploited vulnerabilities in 2026 were hit on or before disclosure day, versus 18.7% in 2018. This is a product announcement with no new incident, but it carries some vendor-reported exploitation statistics.

    Technical Details

    • Agents enumerate subdomains and internet-facing systems (websites, APIs, VPN gateways, mail servers, self-hosted tools such as GitLab). They then attack each host using known CVEs, misconfigurations and exposed secrets.
    • Findings are validated and reproduced by other agents before reporting, and a supervisor agent blocks potentially disruptive actions.
    • Example attack chain: an exposed .git folder leaks repository history, which contains SSH credentials. These work on another host, where Azure credentials are found, giving admin access to the cloud tenant.
    • The article also references CISA's 48-hour disconnect directive for Ivanti VPN appliances in 2024.
    • Aikido's 2026 State of AI in Pentesting report (400 leaders surveyed) says 48% find pentest findings outdated on arrival.

    Remediation

    • Test the external perimeter continuously rather than once or twice a year, and retest after fixes.
    • Inventory internet-facing assets and subdomains, and remove exposed .git folders, secrets and misconfigurations.
    • Patch quickly or take exposed systems offline when critical flaws are disclosed.