Company
Making AI safe to deploy.
AIFenders exists to give every team a runtime firewall for LLMs and agents — so they can ship AI without shipping breaches. One base_url swap puts five declarative rails between any application and its model, enforcing safety on every request and response across the whole lifecycle.
Live product · published benchmarks · open-core · GetDeaddicted LLC
At a glance
- ✓ What we build — a drop-in, OpenAI-compatible guardrail proxy for LLM and agent apps.
- ✓ Stage — live product, published benchmarks, raising a seed round.
- ✓ Model — open-core: OSS engine, hosted cloud, enterprise self-host.
- ✓ Company — GetDeaddicted LLC, engineering-led.
Mission
Runtime security should not be a rewrite.
The teams building with LLMs and agents are moving faster than the controls that govern them. AIFenders exists to close that gap. We believe every team deserves a runtime firewall for their AI — one that enforces policy on every prompt, retrieval, tool call, and streamed token, without a new language to learn or a line of application code to change. When shipping AI safely is a one-line swap instead of a security project, safety becomes the default, not the exception.
Why now
The market is arriving at the layer we sit on.
Three shifts are converging on runtime AI security — and the incumbents are on the wrong side of them.
Agents hit production
Tool-using, multi-step agents are moving from demos to real deployments — the fastest-growing pattern and the one with the largest blast radius when it runs unguarded.
Regulation has arrived
The EU AI Act, India's DPDP Act 2023, and GDPR turn AI data handling and risk management into hard obligations. Buyers now need runtime enforcement and an audit trail.
Spend moves to runtime
The 2025+ security conversation has shifted from scanning the model to controlling agent behavior and egress at request time. Incumbents are mostly prompt/response scanners — not whole-lifecycle proxies.
What we've built
Traction, not a deck.
AIFenders is a live, working product — proxy, dashboard, benchmark harness, and compliance mapping — with numbers we publish in full.
AIFenders is live at aifenders.com with published benchmarks and an auditor-ready mapping to OWASP LLM Top 10, the EU AI Act, and India's DPDP Act — most vendors publish neither their false-positive rate nor their latency. It was originally built as a benchmarked entry for the Crest Data AI Innovation Challenge (LLM Guardrails Middleware) and is now being productized by GetDeaddicted LLC.
Our approach
The principles the product is built on.
- ✓ Developer-native. Policy is declarative JSON — readable, diffable, no Colang or DSL to learn. A platform engineer trials it in an afternoon.
- ✓ Honest. We publish false-positive rates and latency alongside recall, and label coverage truthfully — including where we are not yet at parity.
- ✓ Open-core. A free OSS engine, a hosted multi-tenant cloud, and an enterprise self-host tier — the engine wins the developer, governance closes the account.
- ✓ Privacy-first. Zero runtime dependencies make it trivially self-hostable inside a VPC, and in gateway mode AIFenders takes no custody of provider keys.
Team
Engineering-led, and hiring.
AIFenders is founded and built by the team at GetDeaddicted LLC. We are an engineering-led company, pairing hands-on LLM and agent security work with product and go-to-market experience, and originated the product as a benchmarked entry in an industry AI guardrails challenge before committing to build it into a company.
We are hiring against our roadmap across security research and go-to-market. Detailed founder and team bios are being finalized and will be published here.
Investors
Currently raising a seed round.
AIFenders is at seed stage. The full memorandum covers the problem, the five-rail solution, our technical moat, market framing, competition, business model, use of funds, and roadmap. It is the best place to understand where we are and where the capital goes.
Investor contact: getdeaddictedllc@gmail.com
Let's talk.
Whether you are an engineer who wants to firewall an agent, a design partner deploying customer-facing AI, or an investor sizing up the runtime AI security layer — reach out.