A prompt that once produced calibrated outputs begins to overreach, because the model was upgraded and the instruction was not retuned.
A policy that once excluded sensitive data begins to include it, because a retrieval source was widened upstream.
A tool schema that once restricted side effects begins permitting them, because a new field was added without revisiting the contract.
None of these triggers an alert.
Behavioral Drift as the New Outage
Instruction as Infrastructure
Sovereign AI Infrastructure
Authored by Tharaka Mahabage: Enterprise Cybersecurity Architect with seventeen years across enterprise platforms, government coordination, and higher education. Director, Sri Lanka CERT|CC; board member, Sri Lanka Accreditation Board (SLAB); MSc, Cybersecurity. Visiting lecturer at the University of Kelaniya, CICRA Campus, and NSBM. Mentor, STEM Link Bootcamp.
Security reports: security@tharakamahabage.dev · policy · security.txt · PGP key. Coordinated disclosure is welcome.