AI Infrastructure

We are looking for a solution to help us with identifying an AI solution to help us with end to end delivery of AI use cases. We are begginning our journey when it co.es to managed AI delivery and an associated platform that turns enterprise needs and use cases into real world production ready AI solutions.

Like many companies, right now we have little insight and no control over what generative AI tools that our employees use. While we do have some tools that are covered by formal T&Cs to limit model training, employees are able to use other tools, or sign in with personal users yet still upload or copy and paste proprietary data (user lists, code, etc.). We want to be able to monitor and control this, but our existing toolsets do not offer anything like this currently.

We have multiple sources where knowledge sits. I need to bring everything into serviceknow so AI agents and skills in service now can use that to help users. Task involved gathering information from multiple sources, categorise them, and then create AI friendly unique KB articles.

We are researching tools that let our red and purple teams safely simulate adversarial attacks against AI and LLM-based systems, as well as validate detection and response for AI-driven threats across the broader environment. The ideal platform can emulate techniques such as prompt injection, jailbreaks, model and data exfiltration, insecure tool use, data poisoning, and other tactics mapped to frameworks like MITRE ATLAS and ATT&CK. We want a large, continuously updated library of AI-specific attack scenarios, plus the ability to author and customize our own, and to run them safely against target applications, models, and pipelines without impacting production. Strong purple-team capabilities are important: the tool should correlate each simulated technique with whether our controls detected, blocked, or missed it, then produce clear coverage maps, gap analysis, and prioritized recommendations to improve defenses. We value integration with SIEM, EDR, and detection engineering workflows so results feed directly into tuning and validation cycles. Continuous and scheduled testing, safe execution guardrails, measurable metrics on control effectiveness over time, and executive-ready reporting are all key. Deployment flexibility, strong isolation, and support for both traditional and AI-native attack surfaces matter to us. We intend to evaluate scenario realism, breadth of AI attack coverage, and quality of remediation guidance during a hands-on proof of value.