What’s at stake
A telecom operator runs systems that can’t go down: the network, billing, provisioning and customer care. Behind them are large software teams, a lot of legacy code, and the personal and traffic data of millions of people. In several countries in the region, telecommunications are also considered critical infrastructure, with their own cybersecurity obligations.
AI agents promise the most right there: understanding legacy code, automating tests and migrating systems. But an agent with production credentials is a risk no Security leader will approve.
What you gain with Harkaq
The agent near the system, not inside it. It works in an isolated workspace with no keys and only reaches internal systems through the gateway, with the actions the policy allows.
Control per operation, not per destination. Querying inventory may be allowed; changing network configuration isn’t. Anything new is approved by someone else.
Data that doesn’t leave. The gateway inspects everything that leaves, including what is sent to the model, and blocks whatever it can’t inspect.
Evidence for the sector regulator. Every action goes into a hash-chained log, ready for audit.