supa Debuts AI-Driven Penetration Testing Priced From 500,000 Yen
The service packages the attack-automation techniques that Google Cloud and Anthropic both reported being used against real targets, and sells the defensive version at a price and turnaround that manual testing has not offered.
Reporting from 1 source: ASCII.jp.
supa Inc. has started offering Security Checkup AI, a penetration testing service that reproduces attack methods using frontier and open-weight AI models. It runs on the company's own GPU servers, so vulnerability data never leaves to external clouds. Pricing is 500,000 to 1,200,000 yen before tax, which the company estimates is up to one-seventh the cost of manual expert testing. Reports arrive in as few as three business days.
Security Checkup AI pairs an LLM for security verification with a harness that controls AI-driven tool execution, and white hat hackers help build scenarios from real operations and user privileges. The attacker-role models run on supa's own GPUs, so the systems under diagnosis and any vulnerabilities found are kept off external services. Security engineers check the validity and impact of the results. The service follows a stretch of disclosed attacks: Google Cloud's September 2026 report on automated reconnaissance, and Anthropic's November 2025 disclosure of a campaign that abused Claude Code against roughly 30 organizations.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.