Who Moderates Your Comment Section Now.

TikTok has filed a WARN notice with the Tennessee Department of Labour and Workforce Development covering 250 roles at its Nashville office, which closes permanently on 5 October.

Content moderation is among the work done there. It follows trust and safety reductions in Singapore, Berlin, London and Dublin over the past eighteen months, each of them running alongside a stated shift towards automated first-pass review.

The story most brands should take from this is not a policy story. It is a staffing question about their own comment section, and community management is the job nobody has assigned.

Publishing was never finished when the post went out, and the set-and-forget habit is what leaves a comment section running unattended.

What TikTok filed, and when it takes effect

The notice was filed on 6 August 2026 and covers 250 workers at the office TikTok leased in 2024. The site shuts on 5 October, and the local workforce development rapid response team has been activated to support those affected.

A spokesperson for the TikTok USDS joint venture said the company had “decided to close our Nashville office to streamline our operations and better align our teams for long-term growth”, adding that it remains “fully committed to providing secure, safe and positive experiences for the 200 million Americans” who use the app.

Under the federal WARN Act, employers give 60 days’ notice of a mass layoff. That is why an October closure became public in August.

Nashville is the latest site, not the first

These are separate decisions in separate markets, and they are worth keeping separate:

• February 2025: trust and safety staff cut in Singapore, part of a round that also touched teams in Europe and the US.
• July 2025: TikTok moved to break up the German-language trust and safety team in Berlin, outsourcing the work and replacing part of it with automation. The union ver.di ran a warning strike that month and a four-day strike in September.
• August 2025: several hundred London trust and safety roles put at risk, which TikTok described as “concentrating our operations in fewer locations globally”.
• July 2026: around 300 roles at risk in Dublin, out of more than 2,000 Irish staff, alongside a proposal to create hundreds of new specialist safety roles in the same city.
• August 2026: Nashville, 250 roles, closing 5 October.

Dublin is the useful counterweight. TikTok is not simply subtracting moderators, it is proposing to trade frontline review for a smaller, more specialist layer while automation handles the first pass. That is a reshaping. It still means fewer people looking at things day to day.

What the first pass is good at, and what it is not

Automated review is strong on the obvious: nudity, gore, spam rings, banned terms, known violating audio. Setting out the London changes, TikTok pointed to automation already removing around 85% of rule-breaking posts.

What it handles badly is context. Sarcasm that reads as praise. A furious customer whose complaint is legitimate and whose language is not. Twenty individually mild comments that together are a pile-on. A false claim about your product that a lawyer would call actionable.

Those are judgement calls, and judgement is precisely the layer being made smaller and more specialist.

The risk lands in your comment section

Nobody has published figures showing brand-level moderation response times getting worse, and claiming otherwise would be inventing a number. What can be said is structural: when fewer people handle the cases automation cannot decide, the queue for those cases gets longer rather than shorter.

For a brand, the exposure was never the obvious spam. It is the twelve hours between a nasty thread starting under your best-performing post and anyone at the platform looking at it.

That gap has always existed. It is widening, and it is yours to cover.

Give the comment section a named owner

Ask who owns comments on your accounts and you usually get a function rather than a person. “The social team” is not an owner.

An owner is a name, with hours, and named cover for the hours they are not working. Write down three things:

• Who checks comments, on which accounts, at which points in the day.
• Who covers evenings, weekends, and the fortnight that person is on holiday.
• What “checked” actually means, because scrolling the notifications tab is not the same as reading replies on the three posts currently getting reach.

A rota for publishing is normal. A rota for replying is rarer, and it is the one that matters at nine on a Saturday night.

While you are at it, ask when the keyword filter list on each account was last updated. If nobody can answer, it is sitting at default.

Write a response standard, not a vibe

A response standard says what happens to a comment, and how quickly. Four buckets covers most of it:

1. Reply. Questions, praise worth acknowledging, anything a watching customer would want answered. Same working day.
2. Hide. Slurs, spam, off-topic abuse, anything you would not tolerate in a shop. No debate, no reply, no announcement. Within the shift.
3. Escalate. Legal threats, safety concerns, complaints with substance, anything naming a member of staff. Within the hour, to a named person.
4. Leave alone. Fair criticism, jokes at your expense, mild negativity. Deleting these creates the story the comment did not.

The fourth bucket is the one teams get wrong. Not every comment is a task.

The standard should fit on one page and be usable by somebody covering a shift for the first time. If it needs interpreting, it is not a standard. This is ordinary workflow and operations work, and it belongs next to your publishing process rather than in one person’s head.

The escalation path needs a name at the end of it

Escalation without a decision maker is just forwarding. Decide now, in writing:

• Who rules on anything legal or defamatory, and how they are reached outside office hours.
• Who rules when a comment describes genuine harm or distress, and what the team does in the meantime.
• Who may speak publicly on a live complaint, and who explicitly may not.
• What the community manager is allowed to do without asking, so the routine cases never reach a director.

That last line is the difference between an escalation path and a bottleneck. If everything escalates, nothing does.

Keep a record, and pre-write the holding line

Two habits do most of the work here.

Record before you act. Screenshot the comment and the thread before hiding or deleting anything, then log the time, the account, the action and who took it, somewhere the whole team can find rather than in one person’s inbox. If a complaint becomes a formal one, that record is the entire defence.

Then pre-write your holding lines. The clever reply is the wrong ambition. An approved, on-brand holding line available at 11pm beats a perfect response that lands on Tuesday.

Draft three or four in advance: we are looking into this, please send us your order number; we are aware and will update; this account does not comment on that. Approve them once, alongside the rest of your approval process, and they stop needing approval in the moment.

How NBK thinks about community management

Community management gets treated as the thing someone does between posts. It is a shift, with an owner, a standard, an escalation path and a record, and it deserves the same operating rhythm as publishing.

A brand taking 200 comments a day with no named owner does not have a content problem. It has an unassigned job.

The platforms are reorganising around automation for the first pass and a smaller specialist layer behind it. The sensible response is not alarm about safety policy. It is making sure the person responsible for the second pass on your own account is someone you employ.

Next step

If your social process is leaving the comment section unowned after six o’clock, NBK can help rebuild the workflow behind the content.