Phrase TMS Releases 2026

Project managers with required permissions could not add a term base to a project template. This has been fixed.

The relevant term bases for Norwegian were not showing up when searching by Norwegian language codes, and vice versa. This has been fixed.

Vendors were unable to create jobs in shared projects when custom segmentation rules were used. This has been fixed.

Some APCs were missing history logs on their History page, even though the logs existed. This has been fixed.

The File import settings page for .XML files did not display elements and attributes that had Import selected checked. This has been fixed.

The MT service was shown as pre-selected when editing a project or template, but this choice wasn’t saved after adding a new target language. This has been fixed.

ChatGPT's output sometimes included prompt-like content instead of just machine-translated content. This has been fixed.

Vendor 2 assigned to shared jobs by Vendor 1 could not see the TMs and TBs assigned to the project by Vendor 1. This has been fixed.

Project managers with rights to edit and import in term bases could not edit or import terms. This has been fixed.

The AI Usage analytics dashboard was displaying an error message instead of showing usage data. This has been fixed.

Reimporting a source file from Google Drive set most segments to unconfirmed, even when there were no changes in those segments. This has been fixed.

To enhance data privacy and security, user accounts that are no longer associated with any organization will now be automatically scheduled for deletion after a 30-day grace period.

Users who lose access to all Phrase products are now automatically scheduled for anonymization after 12 months. Platform administrators receive email reminders 7 days after access removal and 30 days before anonymization, while Platform members are not notified. A badge in the Users page indicates accounts scheduled for anonymization.

TM match scoring has been adjusted so asymmetric wrapper-tag matches are downgraded from 100% to 99% fuzzy, preventing unsafe tag structures from being auto-applied during pre-translation.

Some translation memories could not be added to project templates when using certain Norwegian language combinations. This has been fixed.

Download file names showed only underscores instead of the Japanese job title when using Download Target Preview. This has been fixed.

Jobs on the second page of the project were not loading and showed an error message instead. This has been fixed.

New project templates did not accept dynamic naming and showed an error message instead. This has been fixed.

The Translation memories section in the project template wizard was cleared when adding new workflow steps, instead of keeping the existing assignments. This has been fixed.

Clicking a job link from the Recents sidebar opened the project page without pre-filtered job. This has been fixed.

Project managers could not filter projects by Client, and the client list was not displayed as expected. This has been fixed.

New project templates could not be saved unless the Project name field was filled in, unlike the classic view where this was not required. This has been fixed.

The net rate amounts in the analysis view were showing incorrect values when a Net Rate Scheme was assigned. This has been fixed.

When creating a project from a template, removing and re-adding target languages could select the wrong MT profile for those languages. This has been fixed.

Shared jobs showed the wrong step and providers in the assignment template, even though the correct assignment happened in the background. This has been fixed.

Importing Phrase Portal users via file upload was not working, even when trying to import a single user. This has been fixed.

LQA reports were showing the same provider for both translation and review when a job was split in the workflow. This has been fixed.

The APC History log page was very slow to load and sometimes unresponsive, with filters not showing and logs missing. This has been fixed.

Support for the Arabic modern language was missing when using style guides, and some projects with unrecognized language codes could cause errors instead of showing a helpful message. This has been fixed.

Compare analysis from the legacy UI failed when selecting a workflow step to compare. This has been fixed.

Custom field values were not being applied when creating a project from a project template. This has been fixed.

The number of machine translation engines shown in the project settings didn’t match the actual number enabled for each MT profile. This has been fixed.

The next step ready email was sent before the job assigned email, which could be confusing for linguists. This has been fixed.

The Owner and Created by filters in the classic project view weren’t working as expected. This has been fixed.

Workflow settings that were enabled in the classic project template view appeared as disabled in the new project template view. This has been fixed.

The Provider filter in the jobs table was not working as expected. This has been fixed.

Custom project fields were not visible to vendors in shared projects, even though they should have been. This has been fixed.

Translation memories were not showing up in the project template wizard. This has been fixed.

Linguist accounts with the Enable MT option turned off could still use the Pre-translate action. This has been fixed.

Downloaded completed files in Safari on macOS were missing their file extension and appeared as “getCompletedfile” instead of keeping their original name. This has been fixed.

Adding new terms to existing concepts in a term base via .XLSX import was not possible unless the TID column was included, even when it was not needed before. This has been fixed.

Add-ons that were already enabled were still showing up as items to buy in the checkout, making the total price higher than it should be. This has been fixed.

The same column could not be set as both the source and target in multilingual Excel files, which prevented translations from overwriting the original content. This has been fixed.

Some characters were not showing up correctly in .XLSX files. This has been fixed.

Phrase Portal: It is now possible to create a maximum of 21 portals per organization.

Custom field values were not being applied when creating a project from a project template. This has been fixed.

Linguists were seeing a You are not authorized to access the job error when trying to open assigned jobs. This has been fixed.

Some old, unused data for APCs and Services was not being cleaned up from the database as expected. This has been fixed.

Job status was not always updated to EMAILED as expected in the workflow. This has been fixed.

Translating text using Phrase Language API was removing spaces after special character sequences like \r\n in translations. This has been fixed.

Style Guides (Early Access Program)

Language-specific style guides can now be uploaded and managed in the Phrase Platform and attached to projects in TMS, Strings, and Studio. Translators and reviewers can open the original style guide directly in the editor, while MT Optimize and the AI Translation Agent use an AI-optimized representation to follow brand voice, tone, and audience requirements more consistently and help reduce rework.

Quality profiles

Reusable Quality profiles can now be created in the Phrase Platform by combining custom AI checks into a single pass/fail evaluation for each segment. These profiles can be applied in TMS pre-translation settings to automatically lock segments that meet defined quality criteria. Failed checks are displayed in the CAT web editor, helping identify segments that may require review while allowing teams to skip post-editing where appropriate.

Insight alerts are now generated when the AI Translation Agent is estimated to outperform the currently configured MT engines for a specific language pair. These alerts highlight opportunities to improve translation quality and post-editing efficiency based on recent activity data.

.XLSX files with certain regex settings could not be imported, even though they worked before. This has been fixed.

Some tags were missing from completed MS Excel files, even though they appeared correctly in the CAT web editor and preview. This has been fixed.

Import of a bilingual .CSV file was failing after 24 hours due to a timeout issue. This has been fixed.

Certain project jobs were inaccessible in the joined view. This has been fixed.

Unsupported or invalid Excel files could cause processing errors instead of being rejected with clear error messages. This has been fixed.

File names were being cut off at the last slash or backslash when exported, so only part of the name appeared. This has been fixed.

Phrase Language AI was skipping line breaks when using certain translation engines. This has been fixed.

Some analyses with machine translation matches failed for certain languages due to an unsupported DeepL setting. This has been fixed.

Phrase Next GenMT returned prompt-like text instead of a translation during pre-translation. This has been fixed.

Importing multilingual .XLSX files failed with internal errors. This has been fixed.

MT suggestions coming from Mirai (third-party engine) were not showing up in the CAT web editor for the en-us > ja-jp language combination. This has been fixed.

Third-party Microsoft Custom Translator engines were showing Language pair is not supported errors for language pairs that should work, like English to Welsh. This has been fixed.

Missing spaces next to tags and misplaced characters appeared when translating into Korean with Google Translate. This has been fixed.

Running QA checks on projects with a certain term base assigned took extremely long time. This has been fixed.

The file listing in Tridion could sometimes cause an error and fail to load if a file was missing or if there were duplicate files. This has been fixed.

An improved user management experience is now available in the Phrase Platform, enabling Platform Owners and Administrators to manage user access and roles across all Phrase applications from the Platform Users page.

Admins can invite users, assign product-specific roles, and add or remove access to individual applications without affecting other permissions.

APC: Duplicate projects were sometimes created from Tridion Docs connector, even when nothing had changed in the source files. This has been fixed.

Clicking on a job in the search page provided a link with project only, without filters applied. This has been fixed.

Project templates stored term base (TB) information from removed languages. This has been fixed.

After importing certain files, the In-context preview would show duplicated and mixed words in the target text. This has been fixed.

Pre-translation sometimes got stuck and jobs failed to pre-translate. This has been fixed.

UI improvements to the Project templates page:

Portal admins can now display a portal description to end users. When a description is available, a Description button appears in the portal header. Clicking it opens a side panel showing the portal name and description.

Available as of January 30, 2026

Introduced a new profile picture editor in the Phrase Platform profile settings. Users can upload a custom image or generate an avatar with initials and a pictogram. As a result of this update:

APC:

The Submitter Portal was stuck on loading after submission, although the project was created successfully. This has been fixed.

Some projects were not set to Completed even though the project status automation settings were enabled. This has been fixed.

Japanese text was garbled on the Submitter Portal page. This has been fixed.

Importing .DOCX files containing certain inline hidden-text structures was failing. This has been fixed.

CDATA ending tag ]]> inserted in the CAT editor was not displayed in the completed file. This has been fixed.

Portal user could not download a reviewed file due to a 403 error. This has been fixed.

Importing .XLIFF 1.2 files was failing randomly with validation error. This has been fixed.

The Google Translate MT engine was throwing a Bad language error. This has been fixed.

Upcoming changes to data handling in Phrase Analytics and Phrase Data will become effective January 28th. In line with our internal data storage policy, any customer-input text from permanently deleted content does not appear in Phrase Analytics. Permanently deleted content is replaced with Deleted. All numerical values, dates, timestamps, boolean data types from permanently deleted content remain available. This ensures that historical KPIs, trends, and forecasting aggregations remain consistent over time.

The recalculation of analyses containing null job_part in analyse_job were failing. This has been fixed.

Global analysis settings were not being saved. This has been fixed.

It was not possible to open a project template in a new tab. This has been fixed.

The zh-HANT-TW label was obstructing the display of terms. This has been fixed.