# Tharaka Mahabage

> Operational Architecture for the AI Era

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](https://tharakamahabage.dev/essays/behavioral-drift-as-the-new-outage.html)
- [Instruction as Infrastructure](https://tharakamahabage.dev/essays/instruction-as-infrastructure.html)
- [Sovereign AI Infrastructure](https://tharakamahabage.dev/essays/sovereign-ai-infrastructure.html)

## Authorship

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.

## Contact

- Email: hello@tharakamahabage.dev
- LinkedIn: https://www.linkedin.com/in/tharaka-mahabage-bbb14935/
- GitHub: https://github.com/thaaaru
- YouTube: https://www.youtube.com/@TharakaTech42

## Security

- Policy: https://tharakamahabage.dev/security/
- security.txt: https://tharakamahabage.dev/.well-known/security.txt
- PGP key: https://tharakamahabage.dev/.well-known/pgp-key.txt
- Security contact: security@tharakamahabage.dev