Network School announced Thursday it bought Farcaster from Neynar, the platform's second change of hands this year. Users said accounts were restricted the same day, and the new owner writes the main app's rules while promising an open protocol.
Perspective Coverage
4 publishers
- Builder
- Builder 31%
- Operator
- Operator 36%
- Investor
- Investor 33%
Reality
- Evidence55
- Adoption
- Not scored
- Hype gap+30
- Incentives60
- Confidence65
X suspended more than 13,000 accounts it ties to Iranian, Russian and Chinese influence operations before the Nov. 3 midterms. SC Media reported that X did not provide specific evidence for the attributions.
Reality
- Evidence35
- Adoption
- Not scored
- Hype gap+15
- Incentives55
- Confidence55
Ofcom says X met its review-time targets but has not convinced it that illegal hate and terror posts come down. The Grok probe is moving toward a decision under rules that allow fines up to 10% of qualifying worldwide revenue.
Perspective Coverage
3 publishers
- Builder
- Builder 19%
- Operator
- Operator 38%
- Investor
- Investor 43%
Reality
- Evidence55
- Adoption
- Not scored
- Hype gap0
- Incentives
- Not scored
- Confidence70
buildConfirmed2 publishers arXiv now limits each submitter to two papers a month, after a record 40,363 submissions in September. The cap is the latest step in a year of tightening at a preprint server whose volunteer moderators are overwhelmed by thin, often AI-written papers.
Reality
- Evidence74
- Adoption70
- Hype gap+10
- Incentives35
- Confidence76
buildOne report1 publisher Perspective API's shutdown turns error responses into clean 0.0 scores in Discourse's plugin
Google's Perspective API stops working after December 31, 2026. Paxmod, which sells a rival moderation API, says Discourse's official plugin scores an HTTP error as a clean 0.0, so posts would go through unflagged once calls start failing.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+10
- Incentives70
- Confidence45
Harassment of women misidentified as Cornell's Jane Doe runs in the open on X and Substack
Lawyers for Cornell's Jane Doe say three women were falsely identified and harassed by people trying to unmask her. The Verge traces the hunt to X and Substack, mainstream platforms it says are doing nothing to stop it.
Reality
- Evidence55
- Adoption
- Not scored
- Hype gap+10
- Incentives50
- Confidence50
Meta's Oversight Board tells AI labs their new safety committees need the power to cause friction
Meta's Oversight Board issued six recommendations for the safety committees six AI companies, including Google and OpenAI, pledged last week. The accord behind the pledge would fill those committees with company directors, so the independence the board calls essential is in doubt from the start.
Reality
- Evidence55
- Adoption
- Not scored
- Hype gap+5
- Incentives50
- Confidence60
buildConfirmed2 publishers Musubi released PolicyLM-1.7B under Apache-2.0 to score messages against a written policy in a reported 22 milliseconds on an H100. Teams can rewrite rules without retraining, though every speed and accuracy figure so far comes from Musubi's own tests.
Perspective Coverage
3 publishers
- Builder
- Builder 38%
- Operator
- Operator 37%
- Investor
- Investor 25%
Reality
- Evidence45
- Adoption10
- Hype gap+25
- Incentives70
- Confidence60
AI 'comedian' accounts lift working comics' sets to steer Instagram viewers to OnlyFans
Tiffany Sloane, an AI 'comedian' on Instagram, drew about 375,000 followers in 24 days while steering viewers to an OnlyFans page. The working comics whose sets are copied onto these synthetic performers carry the cost of that reach.
Reality
- Evidence50
- Adoption48
- Hype gap+8
- Incentives60
- Confidence45
Meta, TikTok and X are fighting Ofcom in a London court over moderation data demands under a law that allows fines of up to 10% of global turnover. The ruling sets how much the platforms must count and hand over, a week before Meta separately contests how Ofcom calculates fees and penalties.
Perspective Coverage
5 publishers
- Builder
- Builder 23%
- Operator
- Operator 56%
- Investor
- Investor 21%
Reality
- Evidence62
- Adoption
- Not scored
- Hype gap0
- Incentives42
- Confidence68
arXiv limits researchers to two submissions a month to protect its volunteer moderators
arXiv now limits each researcher to two submissions a calendar month and three active at once to curb a flood of AI-written papers. Moderators who once throttled heavy submitters at their own discretion now apply one fixed rule to every account.
Reality
- Evidence58
- Adoption65
- Hype gap+12
- Incentives40
- Confidence60
Facebook health ads borrowing fringe-group wording drew more clicks in 11 of 12 tests
Researchers studying Australian Facebook groups found a vaccine-information ad drew about 52% more clicks when worded as a doubt-raising question. The paired design isolates wording well, though its outcome is the click, so any effect on trust or vaccination remains unmeasured.
Reality
- Evidence50
- Adoption
- Not scored
- Hype gap+25
- Incentives40
- Confidence45
New Mexico asked a state judge for up to $40 billion from Meta after a jury found more than 43 million violations of its Unfair Practices Act. That is seven to eight times Meta's 2019 FTC settlement, sought over public statements about privacy and hate speech.
Perspective Coverage
4 publishers
- Builder
- Builder 10%
- Operator
- Operator 41%
- Investor
- Investor 49%
Reality
- Evidence74
- Adoption
- Not scored
- Hype gap+15
- Incentives70
- Confidence68
New Mexico jurors found Facebook broke the state's Unfair Practices Act nearly 44 million times, at up to $5,000 per violation. A judge sets the penalty in the coming weeks from findings that Facebook's public statements about data sharing misled its users.
Perspective Coverage
5 publishers
- Builder
- Builder 11%
- Operator
- Operator 45%
- Investor
- Investor 44%
Reality
- Evidence72
- Adoption
- Not scored
- Hype gap+20
- Incentives55
- Confidence66
buildOne report1 publisher Ollama's new endpoint puts typed routing and moderation decisions on the local machine
Ollama 0.35 adds a /v1/systemone endpoint returning a choice, yes/no or score from models run on the device. Its 9B Nimble model matched human moderation labels 70.3% of the time in its maker's test, so each team still sets its own review threshold.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+20
- Incentives60
- Confidence45
Six of nine outside advisers quit Big Tech's counterterror group over a Meta-led plan to limit their scrutiny
Six of nine independent experts advising GIFCT, the tech industry's counterterrorism consortium, resigned Monday over a Meta-led plan to curb their scrutiny. GIFCT says the plan is not final. As drafted, the committee could still brief companies on violent trends but could not judge whether their efforts work.
Reality
- Evidence62
- Adoption
- Not scored
- Hype gap+10
- Incentives70
- Confidence58
Santa Fe jurors counted 31 misleading Meta statements on privacy and moderation as about 43 million violations of New Mexico law. Judge Francis Matthew sets the rate on Oct. 1, and at the $5,000 cap the state wants, one statement carries up to $10.5 billion.
Reality
- Evidence72
- Adoption
- Not scored
- Hype gap+40
- Incentives60
- Confidence68
Meta and YouTube reinstated ads for Alex Gibney's 'Musk' on Saturday, a day after a report that four platforms had refused them. For marketers working near politics, YouTube's fix came through its review queue and Meta's came in reply to a reporter.
Perspective Coverage
5 publishers
- Builder
- Builder 11%
- Operator
- Operator 68%
- Investor
- Investor 21%
Reality
- Evidence66
- Adoption
- Not scored
- Hype gap+20
- Incentives55
- Confidence62
Middle schoolers ran a group chat under NPR's Wild Card on Spotify after looking for podcasts with few comments, This American Life reported. Teams that ship comments should expect minors in their least-read threads, where staff are least likely to look.
Reality
- Evidence50
- Adoption
- Not scored
- Hype gap+20
- Incentives
- Not scored
- Confidence55
Reddit withholds suspected spam from the home feed while it reviews it
Reddit says holding suspected spam out of the main feed during review cut home-feed spam by 28% in testing, and it will also stop recommending accounts that were asked to prove a person is behind them and have not done so.
Reality
- Evidence45
- Adoption40
- Hype gap+25
- Incentives78
- Confidence55
Earlier coverage
- The consent fight over AI faces is starting inside families
Leadership · September 21, 2026 · One report1 publisher