AFOKUS

AFOKUS is a cybersecurity company that develops autonomous penetration testing software focused on AI and LLM deployments. Its primary product, AFOKUS Red, is a local-first security assessment platform designed to test AI systems, agents, and LLM deployments without transmitting any data outside the customer's own infrastructure.

The core architectural principle of AFOKUS Red is zero data egress: the tool runs entirely within the customer's network boundary, writes all findings to the customer's own disk, and requires no internet connection, account, or license call during a scan. This design is intended to address data residency, sovereignty, and compliance concerns that arise when using hosted or SaaS-based security scanning tools.

AFOKUS Red covers several assessment areas specific to AI deployments, including:

• Model access and endpoint reachability
• Tool and function surface (parameter schemas, unbounded inputs, execution identity)
• Agent controls (iteration limits, spend limits, loop termination)
• Prompt and context boundaries (system instruction reachability, untrusted input paths)

A distinguishing feature of the platform is its proof-based reporting model. A finding is only reported after the exploitation has been concretely reproduced. Findings are organized into attack paths showing how individual weaknesses chain together toward a target, ranked by how many paths a given fix would close. After remediation, the platform re-runs the original attack steps to verify the fix.

AFOKUS publishes research and framework guidance mapped to standards including OWASP LLM Top 10, MITRE ATLAS, NIST AI RMF, ISO/IEC 42001, and the GCC regulatory layer. The company references GCC (Gulf Cooperation Council) regulatory requirements, suggesting a focus on or presence in the Middle East market.