Limina Launches Contextual Replacement, Bringing Context-Preserving Synthetic Values to Data De-identification
New hybrid capability in release 4.5 replaces sensitive identifiers with contextually accurate synthetic values, keeping de-identified data useful for research, collaboration and AI training
TORONTO, September 14, 2026. Limina, the data de-identification platform built by linguists and machine learning researchers, today announced the launch of Contextual Replacement, a new capability that replaces sensitive identifiers with synthetic values designed to fit their surrounding context. The capability ships in Limina release 4.5 and is included for all customers.
Conventional de-identification approaches protect identity by removing or arbitrarily substituting information, which often strips data of the meaning that makes it useful for key tasks. A hospital may be replaced by an unrelated business, family members may lose a shared surname and records can become incoherent to the researchers, reviewers and AI systems that depend on them. Contextual Replacement addresses this by generating replacements that preserve the character of the original data: organizations stay within their category, related names stay related and structured values such as dates, phone numbers and financial identifiers retain valid, realistic formats.
“Privacy work fails twice when the protected data stops making sense: the organization loses the risk and the value at the same time,” said Patricia Thaine, Founder, CAIO and Chairwoman of Limina. “Contextual Replacement reflects the belief Limina was founded on, that understanding language in context is the only way to protect data without destroying what makes it worth protecting.”
“We built this as a hybrid system on purpose,” said Kory Fong, CTO at Limina. “Language models are the right tool where context determines what a replacement should be, and deterministic rules are the right tool where structure does. Combining them in a single request, with automatic fallback, means teams get contextually accurate output without giving up coverage.”
Contextual Replacement is available through the new GPU Synthetic Enhanced container image and, like all Limina deployments, runs entirely within the customer’s own environment, whether on-premises, in a private cloud or on any major cloud platform. Data never leaves the customer’s control. The capability is opt-in, with Limina’s existing replacement system remaining the default. That way, no data gets left behind.
The launch extends Limina’s platform for identifying and transforming personal, health and payment card information across 50+ entity types and 52+ languages, used by organizations including Providence Health and Boehringer Ingelheim to unlock restricted data for analytics, partnerships and AI development.
Contextual Replacement is available now in Limina release 4.5. More information is available at Limina Contextual Replacement.
366 Adelaide St W, Suite 404, Toronto, ON M5V 1R9
patricia.graciano@getlimina.ai
About Limina
Limina is the data de-identification platform built by linguists and machine learning researchers. It identifies and transforms personal, health and payment card information across 50+ entity types and 52+ languages, deployed entirely within the customer’s environment. Organizations including Providence Health and Boehringer Ingelheim use Limina to unlock restricted data for analytics, partnerships and AI development without compromising privacy or compliance.