Our mistakes, counted by month
Every rule here is read from rbi.org.in. This page counts the mistakes we have published corrections for. It is rebuilt with the site. It counts only what is on the corrections page, because that is the only record of mistakes we have. A mistake nobody has found is not on it. We say that plainly rather than print a rate we cannot prove.
18 corrections published, from 21 August 2026 to 5 September 2026. The site carries 18,631 plain-English points across 401 rule pages today.
By month
| Month | Corrections | Of what kind |
|---|---|---|
| September 2026 | 3 | 1 date or time, 1 who it applies to, 1 a claim beyond the facts |
| August 2026 | 15 | 4 date or time, 3 number, 1 paragraph number, 1 who it applies to, 2 wording, 3 a claim beyond the facts, 1 wrong content on a channel |
A month with none published means we found none, or found none yet. It does not mean there were none.
By kind, all time
| Count | Kind | What it means |
|---|---|---|
| 5 | date or time | A wrong date, day or time, or a stale time shown as current. |
| 3 | number | A wrong count, amount or figure. |
| 1 | paragraph number | A point pointed at the wrong RBI paragraph. |
| 2 | who it applies to | A page named the wrong kind of institution. |
| 2 | wording | A sentence said the wrong thing, or internal notes reached a public page. |
| 4 | a claim beyond the facts | A page claimed more than the facts behind it supported. |
| 0 | label | A wrong Must Read, Good to Know or For information label. |
| 1 | wrong content on a channel | A channel carried content the edition did not approve. |
The kind of each mistake was chosen by the person who wrote the correction. No program guessed it from the words.
Every correction, newest first
| Date | Kind | What was wrong |
|---|---|---|
| 5 Sep 2026 | date or time | Edition 20 said JAIIB registration had closed on 31 August. That was wrong. IIBF has extended it to 6 September. |
| 3 Sep 2026 | a claim beyond the facts | Sign in did not work on the live site. Our own security setting blocked it. |
| 3 Sep 2026 | who it applies to | Two urban co-operative bank pages said they applied to every regulated entity. |
| 31 Aug 2026 | date or time | We named the wrong day for the government bond buyback. |
| 28 Aug 2026 | who it applies to | Four Miscellaneous rulebook pages opened with another institution’s name. |
| 28 Aug 2026 | wording | Six credit risk pages introduced themselves as customer service rules. |
| 28 Aug 2026 | a claim beyond the facts | Our privacy policy said we collect three things we do not collect. |
| 28 Aug 2026 | a claim beyond the facts | Our trust page said no account existed. Accounts existed. |
| 27 Aug 2026 | paragraph number | We printed the wrong paragraph number on the priority sector page, 29 times. |
| 27 Aug 2026 | number | Two pages gave different counts of the old circulars you can look up. |
| 26 Aug 2026 | number | We said 1,323 records had been checked against RBI’s own index. Only 50 had. |
| 26 Aug 2026 | date or time | The home page said we had not checked RBI in nine hours. We had checked four minutes earlier. |
| 26 Aug 2026 | number | Two more pages worked out their own coverage numbers |
| 26 Aug 2026 | date or time | An appointment from January was shown as if it had just happened |
| 26 Aug 2026 | date or time | The library showed dates as compliance deadlines that were not deadlines at all |
| 26 Aug 2026 | a claim beyond the facts | Two claims on this site went further than the source behind them |
| 23 Aug 2026 | wording | An outside audit found internal work-notes and stale data on public pages |
| 21 Aug 2026 | wrong content on a channel | Edition 4 (20 Aug 2026), LinkedIn and Telegram carried the wrong content |
What this page cannot tell you
- Mistakes nobody has found. They are not here. One person reads this site after it publishes; machines check it before.
- RBI documents we missed. Not counted yet. The sources table shows the newest document read from each RBI page.
- A rate. We do not divide 18 by anything. The right bottom number is not known, so no percentage is printed.
The same numbers, for a machine: mistakes.json.