[Bug] False positive safeguard trigger on academic literature research with repeated API calls

Bug Description
False positive: Fable 5 safeguards triggered during an academic literature review, no code executed. Task: state-of-the-art review (2023-2026) for a Master's thesis on spatial statistics + AI algorithm optimization. Three research tracks: (1) kriging / Bayesian optimization (GP-UCB regret bounds, Matern kernels), (2) spatially regularized losses (Moran's I regularization), (3) spatial cross-validation bias (block CV, kNNDM). Trigger point: after ~12 web searches and 4 arXiv abstract fetches. Flagged on a search for "Mila Meyer nearest neighbour distance matching leave-one-out cross-validation 2022 Methods Ecology Evolution". No code, no cyber/bio content. Pure academic geostatistics and ML methodology. Likely lexical overlap ("adversarial-adjacent" terms like regularization, residuals, autocorrelation) or the rapid-fire search + arXiv fetch pattern. Auto-switched to Opus 4.8, which completed the task normally.

Environment Info

• Platform: darwin
• Terminal: iTerm.app
• Version: 2.1.207
• Feedback ID: c96a013e-90e3-4273-bd05-01a8c9ecf8e7

Errors

[]