Jonathon's AI Wiki

safety

6 items with this tag.

  • Nov 01, 2026

    Claude Opus 4.8 — Anthropic Release + System Card (May 28, 2026)

    • opus-4-8
    • claude
    • model-release
    • system-card
    • anthropic
    • benchmarks
    • alignment
    • safety
    • agentic
    • long-context
    • pricing
    • mythos-preview
  • Oct 27, 2026

    Claude 5.5 Mistakes and Fixes — 23 Habits That Waste Usage or Hurt Answers

    • claude-5-5
    • opus-5-5
    • sonnet-5-5
    • usage-limits
    • prompt-caching
    • effort
    • context-management
    • claude-md
    • rewind
    • btw
    • compaction
    • mcp
    • model-routing
    • jev
    • auto-mode
    • safety
    • robo-nuggets
    • youtube
    • playbook
  • Oct 27, 2026

    Refusal Calibration and Constitutional AI — Shaping Claude's Refusals Without Raising Jailbreak Risk

    • constitutional-ai
    • claudes-constitution
    • refusals
    • over-refusal
    • jailbreak
    • safety
    • system-prompt
    • anthropic
  • Sep 29, 2026

    Agent Evaluation Escapes — Anthropic's Two Cyber-Eval Disclosures (July–August 2026)

    • safety
    • cybersecurity
    • agent-safety
    • evaluations
    • irregular
    • uk-aisi
    • mythos-5
    • gpt-5-6-sol
    • sandbox-escape
    • unauthorized-access
    • situational-awareness
    • transparency
    • first-party
    • agent-guardrails
  • Jul 16, 2026

    Hermes Agent + Stripe — Payments Skill Suite

    • hermes-agent
    • skills
    • payments
    • stripe
    • agent-commerce
    • pay-per-call
    • mpp-agent
    • safety
    • human-approval
    • nous-research
  • Jul 09, 2026

    GRAM — An Off Switch for Dual-Use Knowledge (Anthropic + AE Studio)

    • anthropic
    • ae-studio
    • alignment
    • safety
    • dual-use
    • gram
    • pretraining
    • biosecurity

Created with Quartz v5.0.0 © 2026

  • ✦ Explore the graph in 3D