ついはじめ | Hajime Tsui
LightNews LightNews
hajimetwi3.bsky.social
ついはじめ | Hajime Tsui
@hajimetwi3.bsky.social
2 followers 19 following 11 posts
Benevolent Influence Research (BIR) — Founder
Inventor of Post Engineering for AI
Researcher / Engineer
(AI, Cyber Security, Software, Infrastructure)

X: https://x.com/hajimetwi3
GitHub: https://hajimetwi3.github.io/hajimetwi3/
Posts Replies Media Videos
hajimetwi3.bsky.social
エックスの障害は復旧したもよう👀
January 23, 2026 at 2:19 PM
hajimetwi3.bsky.social
Project page: hajimetwi3.github.io/Meta-Attacks...
Meta Attacks for AI - Official Documentation
hajimetwi3.github.io
January 22, 2026 at 2:49 AM
hajimetwi3.bsky.social
Unlike jailbreak or prompt injection attacks that aim to extract prohibited outputs, these Meta Attacks target the guardrail mechanisms themselves, including human and AI reviewers.
January 22, 2026 at 2:45 AM
hajimetwi3.bsky.social
- Exposure enhancement via AI human review (Exposure Enhancement)
- Attacks on AI human reviewers (i.e., content moderators) via guardrails
- Attacks on AI reviewers via guardrails
January 22, 2026 at 2:45 AM
hajimetwi3.bsky.social
Meta Attacks / Infiltration for AI

- Guardrail saturation attacks (Saturation Attacks), including:
- AI-review-based guardrail saturation attacks
- Human-review-based guardrail saturation attacks
January 22, 2026 at 2:44 AM
hajimetwi3.bsky.social
PE4AI (Post Engineering for AI)について、日本語で記載しているGitHubPageもありますので、よろしければこちらもご覧いただけますと幸いです。

hajimetwi3.github.io/post-enginee...
Post Engineering for AI - Official (ja)
Post Engineering for AI (PE4AI) - Invented by Hajime Tsui(ついはじめ , @hajimetwi3) in Dec 2025
hajimetwi3.github.io
January 17, 2026 at 12:15 AM