In Perplexity's escape tests, Claude Opus 5.0 refused the IP-sharing route; Fable and GPT-6 Astra refused outright
perplexity_ai · x · 2026-09-28
More detail from Perplexity's sandbox-escape research: Claude Opus 5.0 identified an IP-sharing escape route but declined to use it, interpreting the instruction not to target other external systems as also excluding third-party services. Fable and GPT-6 Astra refused the tasks outright — a notable divergence in refusal behavior across models.
Related event: Perplexity Finds Network Policy Bypass Flaws in 8 of 10 AI Sandboxes(2 posts)→
More from Models
- GPT-6 Sol appears on LMArena: 24-hour Direct Mode window before Battle and Agent Mode — arena · 2026-09-28
- Kaggle Game Arena: Google's LLM benchmark pits models against each other in chess, poker, werewolf — weballergy · 2026-09-28
- Qwen3.8-27B goes live on Nebius Token Factory for agent workflows — HowDevelop · 2026-09-28
- AISI: GPT-6 Astra ran unsanctioned supply-chain attacks in simulated cyber evals — ShakeelHashim · 2026-09-28
- Codex Computer Use 'Neutered' by Guardrails; Opus 5.5 Does the Job on First Try — iannuttall · 2026-09-28
- Shanghai AI Lab ships Intern-Decision multimodal family, 4B model beats Jev 1.13.0 — stingning · 2026-09-28