The mechanics — how a return is built, transmitted, and secured. The concrete substance behind the product.
Sources: research/SOURCE_INDEX.md — [S20][S21].
PROGRAM_OPPORTUNITIES.md → PO-1.)Tax season is a massive fraud target, so security is not a feature but a foundation:
- Refund fraud / identity theft — criminals file fake returns using stolen identities to steal refunds. Defending against it (identity verification, anomaly detection, IRS IP PINs) is central.
- Account/data protection — the software holds the richest financial data a person has (file 06), so breach risk is existential.
- Accuracy safeguards — error checks and guarantees exist because mistakes carry real penalties (file 03).
A common question in the AI era: if an LLM can build the guided experience, can anyone just ship a tax filer? Building your own return with AI is easy; building a product that files other people's returns is not. Two different things:
1. IRS e-file authorization [S32][S33][S34]
- Apply to become an Authorized IRS e-file Provider (get an EFIN); software developers/transmitters also get an ETIN.
- Pass Assurance Testing (ATS) every year (Pub 1436 for individual MeF) — IRS-defined test returns whose output must match, validated against that year's schemas/business rules; approval is per tax year.
- Conform to the Pub 1345 handbook, pass a suitability check (background/tax-compliance vetting), and repeat for every state you support.
2. Data-security law [S35]
Tax prep is a "financial activity" under GLBA, so the FTC Safeguards Rule + IRS Pub 4557 require a Written Information Security Plan (WISP), a designated Qualified Individual, MFA, AES-256 at rest, TLS 1.2+, annual penetration testing, and an incident-response plan. Penalties up to ~$50,120 per violation, plus loss of PTIN/EFIN.
3. Accuracy liability & trust (Analysis)
You implicitly stand behind the numbers — so errors mean E&O/malpractice exposure, guarantee costs, and consumer-protection law (the "free" saga, file 10). A confidently-wrong LLM is a liability, not a feature.
Two nuances (don't over- or under-state the wall):
- Real but not insurmountable — hundreds of companies are authorized e-file providers (FreeTaxUSA, TaxAct, Cash App Taxes, many small ones). A funded startup can clear these bars.
- There's a shortcut — newcomers can build on an existing authorized transmitter / "MeF-as-a-service" backend and skip becoming their own e-file provider. That lowers the plumbing wall and pushes the real battleground onto trust, liability, and distribution — exactly where Intuit is strong.
The annual, IRS-certified, deadline-bound update is a structural barrier to entry (you can't casually "ship a tax filer") and the reason the season's calendar is inflexible (file 03). It's also where AI-native newcomers face friction: generating tax guidance is easy for an LLM, but becoming an accepted, certified, secure transmitter — and standing behind accuracy — is not. The moat moved: AI collapsed the tax-knowledge layer, but the right to file, the duty to protect data, and accountability for accuracy remain — which is why the durable defense is trust + distribution, not code (see file 11).