Elections officials said Friday that some Baltimore City election day counts have declined because a “human error” caused workers to overreport the number of votes cast in the city by approximately 590 votes.

The state Board of Elections posts regular updates on its website as votes are tabulated. Mail-in ballot counts were expected to rise on Thursday as officials counted them. The unofficial results for election day were not expected to change at all after the state reported 100% of Baltimore precincts filed — but they have.

Election day vote counts were reduced in races for City Council District 11 and District 8, two competitive races that are currently too close to call.

In a Friday news release, the board said the discrepancy was discovered during the standard “100% verification audit of the upload process of the election results.”

After the unofficial results were first automatically uploaded on election night, the board then uploaded the results into “a certified internal network” to verify the unofficial results against log reports from each polling center.

During this internal audit, “it was discovered that the Baltimore City results in the unofficial election night reporting totals did not match the results in the certified internal reporting system,” the release said.

The discrepancy triggered an automatic investigation by both the city and state board. The verification process was performed twice, and investigators discovered that 590 votes were overreported.

“The source of the overreport was identified as human error and resolved,” the statement said. It did not offer specific information about the error.

An independent, third-party firm conducted its own audits of the results, which is available on the board’s website.

“The State Board and Baltimore City Board of Elections are confident that all votes canvassed up to this point have been aggregated and tabulated accurately,” the board’s statement reads. “And the post-election audit of ballot images, conducted after every election in Maryland, will prove as much.”

Reached Thursday night, city Election Director Armstead Jones was unsure where the discrepancies came from. “One memory stick was found when I was downtown yesterday,” he said. “They found it on the supplies piled on the counter.”

Workers “went back and did 100% verification of all sticks,” he added. “That could have changed the numbers.” But, regarding the reduced vote counts in some races, he said: “Nobody should have lost anything.”

Here’s how the new results affect two close council races.

In District 11, officials initially reported that challenger Zac Blanchard received 2,040 votes earned at the polls on election day, outpacing incumbent Councilman Eric Costello’s 1,794 votes. The post-audit results reduced Blanchard’s total by 300 votes and Costello’s by 230.

The new unofficial count published by the state Board of Elections, which also includes early voting and mail-in ballots, showed that Costello had taken an 87-vote lead over Blanchard.

In District 8, the post-audit reduction took 39 votes from the 1,041 that officials first reported Paris Gray earned at the polls on election day. Bilal Ali was reported to originally earn 964 votes;his election day performance was reduced by 25 votes.

According to the new unofficial count, which includes early voting and mail-in ballots, Gray now has a 142-vote lead over Ali.

The gaps between the candidates are due to change again, as about 9,000 more mail-in ballots are counted next week.

Votes were also reduced in citywide races; the changes do not affect the victories for Mayor Brandon Scott, Comptroller Bill Henry, or Zeke Cohen’s win in the primary for City Council president.

In the race for City Council District 10, votes were raised. Councilwoman Phylicia Porter is still more than a thousand votes ahead of challenger Richard Parker.