⚖️🤖 Lawful AI Daily Brief — 2026-05-01

🛠️ Tool Updates

• Claude Code (v2.1.126): claude project purge landed (--dry-run, --all, --interactive) for full project-state cleanup.
• Claude model routing: /model now reads from gateway /v1/models.
• Observability bump: claude_code.skill_activated now logs trigger source (user-slash, proactive, nested-skill).
• Codex CLI (0.128.0): persistent /goal workflows (create/pause/resume/clear), codex update, richer keymaps/status controls.
• Permission posture: --full-auto is being phased out in favor of explicit trust profiles. (Finally, governance with muscles 💪)

💡 Tip of the Day

Use risk-tiered execution profiles so legal workflows stay audit-friendly by default:

profiles:
  legal_review:
    fs: read-only
    network: deny
  drafting:
    fs: workspace-write
    network: allowlist
    allow_domains: [eur-lex.europa.eu]

⚖️ Legal x AI Watch

• Newly active legal/regulation-adjacent repos:
  • ReguNav/app
  • nicholasraimbault/skytale
  • konjoai/squash
• Compliance angle: map agent permission profiles to legal risk tiers and log every escalation event.

📚 Fresh Papers

• APPSI-139: A Parallel Corpus of English Application Privacy Policy Summarization and Interpretation — dataset for clearer privacy-policy summarization and interpretation.
• Exploration Hacking: Can LLMs Learn to Resist RL Training? — investigates strategic behavior during RL post-training.
• Latent Adversarial Detection — activation-level probing for multi-turn attack detection.
• NeocorRAG — evidence-chain RAG to reduce irrelevant retrieval and boost grounded recall.
• Iterative Multimodal RAG for Medical QA — retrieval loop using multimodal evidence.

🔥 Trending Repos

• AutoGPT — ⭐183k — agent workflow platform.
• prompts.chat — ⭐161k — giant prompt library.
• dify — ⭐139k — production platform for agentic workflows.
• langchain — ⭐135k — agent engineering stack.
• hermes-agent — ⭐125k — personalizable agent framework.

🎤 Standup One-Liner

I tightened our AI stack with goal persistence + explicit trust profiles, and lined it up with compliance-friendly guardrails so speed doesn’t outrun auditability.

---

Repo: lawful-ai-staging