Bring Your Own Key Is Now Generally Available

Every security tool built on foundation models raises the same three questions: whose model is doing the reasoning, where's the data flowing, and under whose terms? Most vendors answer those questions for you and don't hand you the controls. Mallory takes a different approach.

Today we're announcing general availability of Bring Your Own Key (BYOK) for every organization. You can now configure the Mallory Agent to run investigations, alerting, and other agent tasks on the latest OpenAI, Anthropic, Gemini, or other models available through OpenRouter, instead of Mallory's own model account. BYOK is available to every user, on every plan. Bring Your Own Model (BYOM), powered by Amazon Bedrock, remains available for enterprise customers.

What the Mallory Agent unlocks

The built-in Mallory Agent is one of the things that makes Mallory an AI-native Threat & Exposure Management platform, not just another dashboard. It powers investigation threads and serves as a workspace for Mallory intelligence: searching assets and entities, generating reports, running tabletop scenarios, and scheduling recurring intelligence delivery, all reasoning over your threat and exposure data on your behalf.

BYOK and BYOM connect that reasoning to the AI provider your organization has already vetted, contracted, and approved, instead of Mallory's own models. Connect your key, and the Mallory Agent runs on it. Your threads, your reports, and your scheduled intelligence deliveries all continue to work exactly as they do today. What changes is which AI account is doing the work underneath.

The AI vendor decision is already made

Most enterprises have already picked an approved AI provider, and that decision came with contract terms, spend commitments, and a security review that took real time to close out. BYOK lets Mallory operate inside that decision instead of adding a second one, so security, legal, and procurement don't have to reopen questions they've already answered.

Configuring the model doesn't trade away what Mallory does. You still get prioritization built on live adversary activity, an agent that reasons across your exposure data, and intelligence delivered through the workflows your team already uses. The model underneath simply answers to your provider relationship instead of Mallory's.

Inference at your cost

Powerful inference keeps getting cheaper, and that trend has held for the last couple of years. You should be able to put your existing enterprise LLM platform to work securing your organization. Configurable inference means you run those models at your cost, with no hidden upcharges. It's built for security and IT leadership who have already standardized on an AI provider and want Mallory operating inside that decision instead of running alongside it.

Existing customers: read Mallory's documentation on LLM and data usage, or reach out to your Mallory contact to enable BYOK for your organization.

New to Mallory? Talk to sales to learn how Mallory fits your environment.