What's New in CoAuthor
AI Content Disclaimers
CoAuthor now shows an AI-generated-content disclaimer in the chat interface across both the web portal and the Word add-in, and carries that disclaimer through to prompt metadata, supporting EU AI Act transparency requirements.
First-Use Compliance Acknowledgment
The first time you open the compliance “About” information, you’re asked to acknowledge it before continuing. On every later view, the dialog can be dismissed normally.
File Integrity Hashing for Veeva Reconciliation
CoAuthor now records dual file hashes (MD5 and SHA-256) at both import and export, so documents can be reliably reconciled against Veeva for Part 11 workflows.
Improvements
The Manage Sources modal now shows the name of the file you’re swapping out, so it’s clear which source you’re replacing.
Generation-time provenance is now preserved inside exported documents, keeping a durable record of how content was produced.
Bug Fixes
Sources & Documents
Source filters now limit actions to the items currently shown, instead of acting on the full, unfiltered list.
Figures in sectional PDF sources now insert intact, instead of fragmenting composite figures or dropping plot images.
Changing the document Type during Create Document no longer discards the file you already uploaded.
Prompts & Editing
Select All, Deselect All, and individual selections in the Prompt Library now respect the active filter and persist correctly as filters change.
The source search bar in the Word add-in no longer autocorrects non-dictionary terms you type.
Switching a prompt from Free Text to Shift Table no longer erases the instructions you’d already entered.
Stability
When a Layar service call fails, CoAuthor now shows a clear error message instead of an endless loading spinner.
What's New in Layar
Improvements
You can now filter documents by their external ID.
The admin portal now lists 10,000 users instead of stopping at the first thousand.
Bug Fixes
The admin portal now reflects each user’s role accurately.
What's New in Curate
“Mark as Good” Approval
Document approval now uses a clearer “Mark as Good” control in place of the previous thumbs rating, making the approval action more obvious.
Bug Fixes
The internal System Data Fabric is no longer shown to users, so it can’t be browsed or have its documents accidentally acted on.
What's New in Certara GPT
Next-Generation Document Parsing
Document ingestion now runs on a new vision-based OCR and parsing pipeline, with a hybrid orchestrator and refreshed indexing. The result is more accurate extraction of tables, figures, and scanned content from the documents you add. The new pipeline is feature-flagged and isolated from the legacy path for a safe rollout.
Vision Model Support
Certara GPT can now send images to external vision-capable models, enabling analysis of image-based content through supported providers.
External Model Providers
You can now connect an external vLLM environment as a model provider, giving you more flexibility in which models power your workspace.
Improvements
Retrieval results can now return custom document metadata alongside each matched chunk, so downstream steps have more context to work with.
Added request trace IDs to backend responses, making it faster to diagnose issues in support cases.
Bug Fixes
Word documents containing complex tables now import successfully instead of failing during processing.
Certain XLSX files that previously failed now ingest correctly.
Document paragraph chunking now applies the configured sizes correctly during ingestion.
Annotating PubMed content no longer times out.
Generating from a document that’s too large to process now returns a clear, helpful error instead of failing opaquely.
Security, Performance & Stability
Removed the legacy Groovy/Grails console from Layar, Cortex, and Connect to reduce the attack surface.
Disabled an internal environment-information endpoint that didn’t need to be exposed.
Patched a known third-party security vulnerability (CVE).
Improved service startup reliability by preventing an internal data file from growing unchecked.
Authentication & Account Management
Unified Sign-In Across Apps — Signing in to Layar, Composer, Codex, and the workflow tools now routes through a single unified login.
Password reset emails now work correctly when Layar is served behind a proxy.
Email verification no longer sends a duplicate confirmation message.
The OAuth token endpoint now returns JSON as expected, instead of XML.