Meta has blamed artificial intelligence tools connected to Anthropic's Claude for the wrongful suspension of the No Kings Coalition's main Instagram account, an error that knocked the protest movement's social media presence offline for hours during a nationwide day of action aimed at boosting early voter turnout ahead of November's midterm elections.
The suspension, which began Saturday afternoon, was first attributed to routine automated protections. A day later, Meta offered a different and more unusual account: according to reporting from Bloomberg and The Guardian, the company's engineers found that tools connected to Anthropic's Claude had been pulling posts, profiles and hashtags from the account at roughly four times the usual pace — traffic patterns that Meta's systems mistook for automated bot activity and responded to by taking the account down. For more context on this story, see our ongoing latest AI developments.
Hours of Silence During a National Day of Action
The lockout could hardly have come at a more sensitive moment. On Saturday, organizers staged nearly 3,000 "No Kings: Vote Early" events across the United States, according to figures cited by The Deep Dive, part of the pro-democracy movement that has drawn millions of Americans to the streets in recent years.
Administrators lost access to the account at about 1:30 p.m. Eastern, a coalition spokesperson said. The account itself reappeared publicly after roughly 90 minutes, but the people behind it remained locked out of its backend for close to four hours, regaining full access at 5:33 p.m. Eastern.
Ezra Levin, co-founder of the organizing group Indivisible, described the situation bluntly. "They suspended it. We didn't have access to it," he said.
Meta's Explanation Changed — Twice
Meta's first public response came Saturday, when spokesperson Andy Stone wrote on X that the platform's automated systems sometimes act "when accounts post, share or comment at very high frequencies." He acknowledged that "that automated action was clearly taken in error," adding, "Everything is now working as it should and we sincerely apologize."
By Sunday, the company had revised its account of events. In a longer statement, Stone said a high volume of activity reading profiles, posts and hashtags in "intense bursts" had caused Meta's AI tool to classify the traffic as automated bots. "This activity caused our automated protections to incorrectly take the No Kings Instagram account offline," Stone said, noting that an automated re-review an hour later recognized the mistake and restored the account, though untangling admin access took additional time.
The Claude connection was first reported by Bloomberg on Sunday. What Meta has not explained is who was operating the tools, or why they were querying the account so aggressively. The company did not respond to that question in its public statements.
Organizers Call the Answer Inadequate
The coalition was unconvinced by the bot-traffic explanation. In a post on Instagram Sunday, the group said its members had "posted only a handful of times and livestreamed to thousands" during the day of action — hardly the profile of an automated spam operation.
"We have regained access to our accounts, but Meta still owes us an explanation," the No Kings Coalition wrote. "Restricting communications about voting in the middle of an election is unacceptable."
The episode quickly drew political reaction. Senator Bernie Sanders suggested on X that "billionaire ownership of the media" was behind the suspension, while Connecticut Senator Chris Murphy called for an investigation into what happened.
When AI Traffic Looks Like Abuse
Beyond the immediate politics, the incident illustrates a growing fault line in how large platforms police their systems. Automated enforcement depends on classifiers that distinguish human behavior from bot behavior — a distinction that is getting harder to sustain as AI agents, scrapers and third-party tools query social platforms at machine speeds for legitimate reasons.
Meta's account of the suspension suggests its defenses worked as designed against traffic that genuinely looked automated, and failed in exactly the way critics of automated moderation have long warned: quickly, opaquely, and against the wrong target, with the correction arriving only after the harm was done.
For Anthropic, the episode is a reminder that a model's footprint extends well beyond its own products. Claude is accessed by third-party applications through Anthropic's API, and as reporting on the incident indicates, Meta pointed to tools "connected to" the model rather than anything Anthropic itself ran. Downstream usage, in other words, can put a model maker's name in an unwelcome headline through no direct action of its own.
What Happens Next
Meta has apologized twice and restored the account, but the coalition is demanding a fuller accounting, and the company has so far declined to name the operators of the Claude-connected tools or detail what safeguards, if any, apply to political organizations during election season.
With the midweeks approaching and early voting underway in several states, platform groups and election officials will be watching whether Meta — and other platforms that rely on similar automated defenses — adjust their thresholds to avoid entangling civic organizations in bot filters. Whether regulators take a closer interest may depend on how satisfactorily Meta answers the question organizers keep asking: what exactly happened, and who was behind the traffic that triggered it all.
Reporting from The Guardian, Bloomberg and The Deep Dive contributed to this article.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
