Technology & Security
Technology built to protect children, not to collect them
AIP’s technology is run by its Chief Technology & AI Security Officer and underwritten by Victorino LLC, so donated and granted funds go to programs instead of IT overhead.
Who powers it
Underwritten technology
Operating principles
Our zero-surveillance charter
- AIP does not collect, store, or hoard children’s chats, images, or tip content. Reports go to NCMEC and law enforcement.
- This website has no user accounts, no advertising trackers, and stores nothing typed into its forms.
- Nothing that involves a minor is sent to an outside AI service. Text flagged as involving a minor is blocked, not sent, and is never logged.
- AI never makes an accusation. Any finding is reviewed and signed by a named human before it is used.
- Tools that touch evidence are open to inspection and run offline wherever possible.
Evidentiary verification
SUBROSA // OS
SUBROSA // OS is an open-source evidence workbench designed to run locally on a machine the reviewing team controls. It organizes records, checks them against their sources, and produces a source-linked brief for a human examiner. AIP uses this approach so that material relating to child exploitation is verified under strict handling rules, with a detached, human-signed review.
How it works, in four moves
- 01 · GatherOne sealed case file, on a machine the reviewer controls. Nothing is uploaded.
- 02 · TraceEvery claim links to the record it came from. No source, no claim.
- 03 · ChallengeCompeting explanations are weighed side by side.
- 04 · SignThe software never accuses anyone. A named human examiner decides.
This is a simulation. The files, hashes and findings are invented for this page. It illustrates the method; it is not real software output or real case data, and no child’s information is shown.
Read the run as text
- SUBROSA // OS · offline workbench · SIMULATION
- $ network --status
- air-gapped. uploads disabled. nothing leaves this machine.
- $ case open DEMO-0001 --fictional
- ingest screenshot_threat.png
- ingest chat_export.txt
- ingest account_details.pdf
- ingest payment_request.jpg
- sealing originals read-only, hashing…
- sha256 9f2c71ab04e8d3b6c15a7e90f4d2b8a31c6e5f07a9d4b2e8c3f1a60d7b5e94c2
- $ brief draft --source-linked
- claim 1 threat message, Mar 3 9:41 pm → chat_export.txt:212 ✔ source
- claim 2 $200 in gift cards demanded → payment_request.jpg ✔ source
- claim 3 “the sender is a classmate” → no record found ✘ STRUCK
- claim 4 account made 2 days before contact → account_details.pdf:1 ✔ source
- $ hypotheses --compare
- H1 stranger posing as a teen ████████░░ 3 supports
- H2 someone the child knows ███░░░░░░░ 1 support
- both remain open. software draws no conclusion.
- $ brief submit --to examiner
- STATUS HELD · AWAITING HUMAN EXAMINER
- ✔ reviewed and signed by a named human examiner
Status
What is live and what is in development
Live today: this website (static, no data collection), AIP’s Google Workspace email and file storage, and the free parent and educator toolkit.
In development: automated threat-to-curriculum publishing for new online risks, a policy gateway for any AI use, and an examiner review kit for the SUBROSA // OS bridge. These will be described here as they go into use.
Questions about AIP’s technology or security: joey@aipnv.org. To report a vulnerability in this site, use the same address.
Safety is built before the crisis
Pick your part and see it through. Prevention works when people follow through.