From the source
Sakana AI is launching the Security Research Group, an independent R&D team that studies collective, AI-driven cyber defense and the security and safety of AI itself.
- The group sits apart from Sakana's commercial business and is meant to serve the public interest by publishing research, reproducible benchmarks and practical guidelines, and by working with government, industry and academia.
- Sakana points to a 263% rise in CVE submissions between 2020 and 2025, reported by NIST, and to early 2026 running nearly a third above the same period a year earlier.
- The company argues that faster triage and patching will not bring overall risk down, because attackers keep exploiting well-known weaknesses such as input validation and memory safety flaws.
- It says the lasting answer lies upstream in the development process, keeping flaws from arising at the design and build stage rather than chasing them after release.
- Sakana frames security as must-be quality, which goes unnoticed when present and causes strong dissatisfaction only when missing, and says Japan's quality tradition could let the country lead here.


