How AI scans 50+ years of Tamil Nadu EC entries for unreleased mortgages, stray attachments and extent jumps, and what to do when a flag is real.
Five patterns account for most EC-borne risk. A mortgage or charge created decades ago with no matching release deed. A court attachment or lis pendens entry that was never vacated. A transfer by a party who never appears as a transferee, suggesting a missing link or an unregistered step. An extent that grows or shrinks between entries without a partition or sale to explain it. And gaps, years where a known transaction should appear but does not.
Each is individually rare. Across a 50-year window on a multi-survey-number aggregation, the odds of at least one rise sharply, which is why long-window reads matter.
Once the EC is structured, every entry a row with parties, date, type and amount, anomaly detection is largely deterministic: match every charge to a release, every transferor to a prior transferee, every extent to a deed that explains it. Entries that fail to reconcile are flagged with their document number, year and SRO, graded by severity, and queued for human review.
The model adds value on the messy edges, name variants across Tamil and English spellings, clerical errors in survey numbers, entries indexed against the wrong field, where fuzzy matching catches links a literal comparison would miss.
A confirmed anomaly becomes a condition, not necessarily a dead deal. An unreleased mortgage is cured with a registered release or discharge proof. A missing link is repaired with the intervening deed or a rectification. An attachment requires the court record and, often, a renegotiation. The point of finding it early is that cures take weeks, and finding them after agreement of terms costs leverage as well as time.
In our verification, anomaly findings carry the severity grade, the cure path and the evidence required, so the transaction team negotiates from the record rather than around it.
An entry in the encumbrance certificate that breaks the parcel's expected pattern: a charge without a release, a transferor with no prior title, an unexplained extent change, an attachment never vacated, or a gap where a transaction should appear.
Because the riskiest anomalies are old. An unreleased 1980s mortgage or a decades-old partition still binds the title today, and a 13-year window simply never sees it.
Often, yes: releases, rectification deeds, succession documents or court orders cure many flags. The grade tells you whether to cure, reprice or walk away, and finding it before terms preserves your leverage.