Inverity

The Truth in Every Pixel

Inverity compresses media with neural codecs. That means your assets pass through our models, and you deserve a plain answer about what happens when they do. This page is that answer.


Your media is never training data

We do not train, fine-tune, or evaluate our models on customer media. Ever. Assets you send through Inverity are processed, verified, delivered, and then deleted on the schedule described below. Our models are trained exclusively on licensed and open datasets curated by us.

This is not a configurable setting or an enterprise upsell. It is the default for every tier, including the free one.

How your assets move through the platform

Processing follows a single, consistent path regardless of where your media originates:

      1. Ingest. Assets arrive over TLS from your connected system or direct upload, and are screened for malware on arrival.
      2. Orchestration. The Neural Media Orchestrator routes each asset to the most suitable codec family based on content type and delivery target.
      3. Verification. Every output is scored by the Perceptual Hypercube, our quality verification layer, before it is released. Any output that fails verification is not shipped; the asset falls back to a verified encode instead, so nothing is ever delivered unverified and nothing gets stuck.
      4. Delivery. Verified assets return to your system of record. We do not retain a parallel media library.

Retention

Source assets and intermediate artifacts are retained only as long as needed to complete processing and verification, and are deleted no later than 30 days after processing. In practice, most are removed shortly after delivery; 30 days is the outer limit, not the norm.

Processing metadata (file sizes, codec selections, quality scores, token usage) is retained for billing, debugging, and account reporting. Metadata contains no media content.

If you delete your account, associated metadata is purged within 30 days.

The single exception to these deletion schedules is material we are legally required to preserve, such as apparent CSAM identified under our child-safety obligations described below. Nothing else overrides the schedule above.

Quality governance, not quality promises

Most compression vendors ask you to trust that outputs look fine. We verify it instead.

The Perceptual Hypercube scores every output before delivery. Verification runs on every tier and every asset, not as a premium feature. If an output does not meet threshold, it does not ship, and the asset falls back to a verified encode. This is what "governed and verified" means in practice: a gate, not a dashboard.

Performance claims, methodology, and research live in our benchmarks catalog, published with anchors, datasets, and test conditions.

Model provenance

Our codecs are proprietary models developed and maintained by Inverity. We do not send your media to third-party AI providers for processing. Inference runs on infrastructure we control.

Infrastructure

Inverity runs on AWS, which is our only infrastructure subprocessor. No other party processes your media. Data is encrypted in transit (TLS 1.2+) and at rest. Access to production systems is restricted and logged. Uploaded files are scanned for malware before processing.

Child safety

Inverity has zero tolerance for child sexual abuse material (CSAM). This is a firm line, and we enforce it.

Media processed through Inverity is screened against known-CSAM hash databases. When we identify apparent CSAM, we act on it: we preserve the relevant material as required, report to the National Center for Missing and Exploited Children (NCMEC) in accordance with our legal obligations, terminate the account, and cooperate with law enforcement.

This screening runs alongside the processing described above and does not change how your media is handled in any other respect. Legitimate media is processed, verified, delivered, and deleted exactly as described on this page. Detection here is narrow and specific: it matches against known illegal material, and is not a general inspection of your content.

We treat this as a non-negotiable responsibility of operating a media platform, not a feature.

What we do not do

      • We do not train on your media.
      • We do not sell or share your media or metadata.
      • We do not retain your assets beyond processing and verification.
      • We do not ship unverified outputs.
      • We do not tolerate CSAM, and we report it.

Questions

Security or procurement questions we have not answered here: contact us at security@inverity.ai and we will respond directly.

Last updated: July 2026