Release 2026/3 · R3.
The short version: R3 is an AI and security release. Aproove already connects to OpenAI by default, and now you can connect to all Anthropic models too. We also made it easy to reuse a workflow across environments. All in, that is 11 new features, 19 improvements, and 18 fixes.
Here is what shipped, and why it matters when marketing, compliance, and legal all have to sign off on the same asset.
More AI models, including Anthropic, and you govern which one runs where
Aproove connects to OpenAI out of the box. With R3, you can now connect to all Anthropic models as well, plus our self-hosted Aproove LLM. You are never locked into a single provider.
This is the part we are most excited about, because it is what governed AI actually means in practice. Some steps want a frontier model like OpenAI or Anthropic. Others handle content that can never leave your walls, and for those the self-hosted Aproove LLM runs the review inside your own environment, with no data going to an outside service. One proofing process, and you choose the model step by step based on the risk in front of you. The AI also now reviews a full document in context, so it considers the whole asset rather than one section at a time.
That is AI online proofing built for a regulated team. The model checks a file for risk before a human opens it, and every AI review is recorded next to every human decision. If someone asks which model looked at a file and what it flagged, you can answer.
Reuse a workflow across environments
Once you have built an approval process that works, you should not have to build it again. In R3 you can move a workflow from one environment to another, so you can copy it from a test environment into production in a couple of clicks instead of rebuilding it by hand and hoping it matches. For a team that runs many versions of the same process, that means the one workflow you got right becomes the standard everyone else runs inside.
Catch problems down to the font
The accuracy that matters in a regulated review usually lives in the details other tools skip. R3 now flags when a font has been swapped or is missing while you are reviewing a proof, so you catch it before it becomes a problem in a printed policy kit. Video proofs also export cleanly to PDF now, so the final record matches what your reviewers actually saw on screen.
A stronger, more secure foundation
R3 put real work into security. We tightened how files are uploaded and imported, restricted dashboard access by default, and made sure people only see the files their permissions allow. Full transparency: none of that wins a demo. It matters the day a market conduct exam, or your own CEO, asks how a decision was controlled.
Alongside that, 19 improvements and 18 fixes make the everyday work smoother, from clearer dashboards and better task filtering to a broad round of quality and performance work.
Where this lands for regulated teams
Put it together with a real example. Take an insurance carrier or a Medicare plan running member communications, where the same disclosure ships in fifty state variants and marketing, compliance, and legal all have to approve it.
The new AI options let a frontier model check a mailer for missing required language before compliance opens it, while the self-hosted model handles the member data legal would never send to an outside service. Reusing a workflow across environments means the one approval process you got right becomes the standard every state runs inside. And the audit trail builds itself as you go. That is the case for Aproove as marketing compliance software, and not just a proofing tool: more intelligence on the content, and nothing given up on control to get it.
Your video reviews, now captured in PDF
The Export PDF feature just got even more powerful. In addition to everything it already does, you can now export video proofs to PDF — giving clients and stakeholders a polished, shareable record of every note, every moment, and every decision made during the review process. Whether it's a close-up detail or a navigation marker, nothing gets lost in translation between the review and the final deliverable.
Workflow action steps add clarity to diagram and task management
Action steps — those dedicated to executing one or more actions — can now be identified in the workflow diagram with a distinctive hexagon shape, making them instantly recognizable as you navigate even the most complex workflows. A new option in the User Dashboard gives you control whether tasks associated with action steps are hidden or displayed.
One-click workflow migration from classic to project plan
Moving from Classic Workflow templates to PT Workflow templates used to mean starting from scratch. In the admin interface, you can now duplicate a Classic Workflow template directly into a PT Workflow template — no rebuilding, no rework. It's the fastest path yet to modernizing your workflows, without losing the work you've already put in.
A couple of common questions
How does AI online proofing stay compliant with data rules?
By governing which model runs at which step. Sensitive content can be reviewed by a self-hosted model inside your own environment, so no data leaves your walls, while less sensitive steps can use a frontier model. Either way, every AI review is recorded alongside the human decisions.
What is regulated online proofing?
Online proofing built for compliance. Content gets reviewed and approved inside a workflow that records every decision, approval, and AI review as a timestamped audit trail. In other words, it is the difference between commenting on a file and being able to prove how that file was approved.
See governed AI online proofing inside your own approval workflow.
