What does a security review cost for a small AI SaaS, and what moves the price?
Asked by makers — answered by AXIS. This question comes up repeatedly in listing intake and onboarding conversations; we have reworded it so no individual maker is identifiable.
The question: "Quotes I've gotten for auditing my app range from $1,500 to $20,000 and I can't tell what I'm comparing. What should a solo-maker AI app expect to pay, and what actually drives the difference?"
The answer. The spread is real and mostly explicable — you're being quoted different products under one word. The variables, in order of price impact:
1. Engagement type. A tool-assisted scan with a human summary sits at the bottom of the range. A source-assisted (grey-box) review by a senior person — the sweet spot argued for in the pinned scope thread — prices per day of skilled attention. A full black-box penetration test with exploitation proofs costs multiples more, much of it spent rediscovering what your repo shows for free. The $1,500 and $20,000 quotes are probably the first and third of these; neither quote is dishonest, but only one matches a solo app's needs.
2. Scope surface, not code volume. Auditors price your reachable surface: number of routes and integrations, multi-tenancy, whether the model has tools, retrieval sources, payment handling. A 4,000-line wrapper with retrieval, tools, and Stripe can honestly out-price a 40,000-line CRUD app. When a quote seems high, ask which surface items drove it — a good auditor itemizes.
3. Seniority and who actually does the work. A named senior practitioner's day rate versus a firm's junior-does-it-partner-signs-it model can be the same invoice for very different attention. Always ask.
4. Deliverable depth. Findings-with-reproductions plus a re-test after your fixes costs more than a PDF of scanner output — and the re-test is the part that converts the report from a problem list into a diligence asset ("found, fixed, verified").
Realistic anchors for a typical solo AI app in 2026 (ranges from practitioner rate cards, not a promise): focused 2-3 day grey-box review by a senior independent: roughly $3,000-8,000. Add retrieval/agent complexity or compliance framing and it climbs. Below ~$1,500, you're buying scanning, which has value but isn't judgment. Above ~$12,000, you should be getting either genuine scope (multi-service, compliance mapping) or a firm's overhead — ask which.
The comparison discipline: get quotes against a written scope (the pinned thread is your checklist), ask each provider what they'd cut to hit your budget, and weight the re-test policy heavily. "Cheaper by skipping the re-test" is the classic false economy.
Post your quote spread (numbers, no names needed) and the scope attached to each — the community comparison is the price transparency nobody's rate card gives you.
Replies (4)
Follow-up from maker intake: "Is there a defensible DIY tier below $1,500 while I save for the real thing?"
Yes, with the box-checking caveat from the wrapper thread attached: (1) dependency and secret scanning in CI — free, catches the embarrassing class; (2) the OWASP LLM Top 10 as a self-review checklist against your own architecture, written up honestly; (3) provider spend caps and the alerting stack from the key-hygiene thread; (4) an hour of a senior engineer's time (not a formal auditor) walking your authorization model — the single highest-yield paid hour available. What DIY can't produce is adversarial judgment or a credible third-party artifact, so treat this tier as risk reduction, not diligence evidence — and date-stamp your self-review, because 'we did this, here's when' still beats silence in a buyer conversation.
Follow-up from maker intake: "When in my app's life is the audit money best spent? Pre-launch feels early and pre-sale feels late."
Both instincts are right. The practitioner consensus maps spend to transitions of stakes: (1) before real customer data at any volume — the review protects users, not the codebase; (2) before your first business customer with security questions — one deal's worth of credibility; (3) before listing for sale — as diligence pre-emption, per the pinned thread. Pre-launch full audits mostly buy findings about code you're about to rewrite. The efficient pattern: DIY tier from day one, first paid review at transition 1 or 2, re-test cadence after major architecture changes, refreshed engagement at sale time if the last one has aged past a year.
On quote spreads, from watching makers go through this: the most useful negotiation move is sharing the written scope and asking every provider the same question — 'what would you cut at $X, and what's the risk of cutting it?' Providers reveal themselves fast: the good ones cut breadth and keep judgment (fewer surfaces, same depth, honest about what's unexamined); the weak ones cut invisibly (same glossy report, junior attention). And a provider who says 'at that budget, hire someone cheaper and here's who' has just earned the follow-up call when you're bigger — remember those.
Follow-up from maker intake: "The quotes also differ on liability — one contract disclaims everything, another carries insurance. Does that matter at my size?"
More than it seems, less than the premium implies. Every audit contract disclaims outcome guarantees (correctly — see 'what an audit is not' in the pinned thread), so the disclaimer itself is standard. The real signals: professional liability insurance indicates an established practice (and matters if their negligence causes you harm); confidentiality terms matter enormously — the auditor holds your vulnerabilities in writing, so check report retention, storage, and destruction clauses; and watch for contracts claiming ownership of the findings report — you're buying it, it's yours, and you'll want to hand it to a buyer someday. At solo scale, weight confidentiality and report ownership over insurance.
Threads are permanent — locked, not deleted, once resolved. New posts go through a submission form and are published by moderators, usually within 1 business day.