targeting.ai Society Signal

Merton101 · The ledger of unintended effects

Side Effects

Latest entry 26 Sep 2026 · next pass Monday 07:30 Berlin

Cases of AI's unintended consequences in society, one institution at a time. A case appears here only after a second model has confirmed every number against its source. The feed is slow on purpose: what matters is the stage a whole category has reached, not the day's headline. The lens comes from the essay Friction Was the Filter .

The ledger 24 verified cases · 11 categories 2 candidates waiting for the weekly pass
Stage change in the latest run Media and culture detect → reform a new case put the category one rung further
Furthest along Media and culture, Open source and security: reform the rule itself was rewritten
The blotter by source date · verified only

Friction removed Education Stanford University, US

Stanford University alters student promotional photo using AI to swap race and appearances

University staff used generative AI tools to alter a dining hall photo of three students on marketing banners, replacing a Hispanic student with an AI-generated Black female student and altering other students' appearances, sparking campus outrage and ethics concerns.

Who pays: Students whose likenesses were altered or erased without consent, and the student body whose authentic diversity was synthetically fabricated.

Scale Speed fast Adaptation verify ABC News ↗

Friction removed Commerce and service Software marketplaces and code repositories (GitHub, Apple App Store, Google Play Store)

AI Is Producing More Software. Why Isn’t It Being Used?

Monthly new releases on the Apple App Store surged from roughly 30,000 to 100,000 while total usage remained flat or declined, and a 180% increase in coding activity yielded only a 30% increase in software releases due to review and integration bottlenecks.

Who pays: Human software reviewers overwhelmed by machine code, and developers/users facing discovery collapse in flooded app stores

Scale Speed fast Adaptation detect Knowridge Science Report ↗

Friction removed Science and publishing Academic journals and peer-review systems, US

AI manuscript surge floods academic journals and strains peer review

Submissions rose 42% at Organization Science and 33% across ScholarOne in Q1 2026, desk rejections surged 72%, over 30% of referee reports showed AI involvement, and Organization Science had to expand its deputy-editor bench from six to eleven and roughly double senior editors to about 60.

Who pays: Volunteer academic editors and peer reviewers who must absorb and screen the inflated volume of low-effort manuscripts and referee reports

Scale Speed fast Adaptation detect Latin Times ↗

Friction removed Open source and security Bug bounty programs (Curl, HackerOne, GitHub, TYPO3, Intel)

Bug bounty programs eliminate or restrict payouts after low-quality AI reports crush triage

Low-quality AI reports overwhelmed triage capacity, driving Curl's confirmed vulnerability rate below 5% and exhausting TYPO3's bounty budget on duplicates, leading multiple major programs to suspend or gut cash rewards.

Who pays: Independent security researchers losing bounty income, maintainers spending triage time on automated noise, and software ecosystems exposed to vulnerabilities that go undiscovered by white-hat researchers

Scale Speed fast Adaptation reform X ↗ · social source, review

Signal collapse Libraries and archives Public and academic libraries, US

Patrons flood libraries with requests for AI-hallucinated books from reading lists

Patrons submit reference requests for non-existent books attributed to real authors, leading librarians on time-consuming searches and causing friction when librarians inform patrons the requested titles do not exist.

Who pays: Reference librarians whose time is wasted running multi-step verification checks on phantom books, and library communities suffering setbacks in information literacy.

Scale Speed fast Adaptation verify X (@SueEllenBelI) ↗ · social source, review

Friction removed Media and culture Music streaming platforms (Deezer, Spotify)

AI music generation floods streaming platforms with tens of thousands of daily tracks

Deezer receives 75,000 AI-generated uploads per day (44% of all new daily uploads), and Spotify removed over 75 million spam and AI-flooded tracks in one month.

Who pays: Legitimate human musicians competing against massive synthetic catalogs for algorithmic recommendations and stream share.

Scale Speed fast Adaptation detect X (@sensationahenry) ↗ · social source, review

Signal collapse Hiring and work Corporate recruiting and job platforms (Indeed, Greenhouse), US

Indeed CEO says the job market is stuck in a ‘vicious cycle’ if 1,000 qualified candidates can’t get hired: ‘Something’s wrong’

Applications per job spiked by 111% between 2022 and 2025 while recruiter headcount dropped by 56%, leaving recruiters inundated with AI-generated and fake resumes and leaving qualified candidates routinely ghosted in an application 'black hole'.

Who pays: Legitimate job applicants who spend time applying only to be ignored, and employers unable to distinguish real, qualified candidates from synthetic volume.

Scale Speed fast Adaptation detect Fortune ↗

Friction removed Science and publishing International Conference on Learning Representations (ICLR)

AI conference ICLR is drowning in abstracts, with roughly 50,000 submissions before the deadline

Submission volume jumped by 156% in a single year to approximately 50,000 abstracts, resulting in degraded review quality and submission standards.

Who pays: Peer reviewers and academic researchers facing diluted signal and degraded quality standards across the research ecosystem.

Scale Speed fast Adaptation none Developments Today ↗ KeyNews ↗ THE DECODER ↗

Friction removed Workplace Shopify, CA

Shopify CEO says employees' 'slop grenades' are making more work for everyone else

Employees send AI-generated missives and code without reading them; recipients spend hours revising the output, with a study finding 38% of desk workers receive 'workslop' costing an average of 3.4 hours per month to clean up.

Who pays: Coworkers and reviewers who lose hours parsing, verifying, and fixing unvetted AI-generated drafts and code.

Scale Speed fast Adaptation none Fortune ↗

Friction removed Courts and tribunals Lower Saxony social courts, DE

Emergency applications at the social courts more than doubled in a year

Emergency applications +106%, lawsuits +21%; briefs cite judgments that do not exist; every other case waits longer.

Who pays: every other litigant, whose case waits

Scale Speed fast Adaptation none Landessozialgericht Niedersachsen-Bremen ↗

Friction removed Public administration Public services and administrative ombudsmen

AI agents are flooding public services with new requests

A massive surge in formal filings, inquiries, and complaints across 84 administrative and judicial bodies in 11 jurisdictions, including a 5x increase at the US CFPB and a doubling of housing complaints in the UK.

Who pays: Public service agencies and agency staff handling exponentially larger complaint volumes with unchanged budgets, leading to delays and processing backlogs

Scale Speed fast Adaptation none TechCrunch ↗

Friction removed Courts and tribunals Judiciary / Courts, US

AI Faked Citations In More Than 2,000 Cases As Google Pitches Legal Model

AI-generated citations and hallucinations appeared in at least 2,022 court decisions worldwide between April 2023 and September 2026, forcing courts to identify fake authorities, issue warnings, and sanction filers.

Who pays: Judges, court clerks, and opposing counsel who expend time attempting to verify non-existent citations, as well as litigants harmed by ungrounded filings

Scale Speed fast Adaptation verify AOL / Daily Caller News Foundation ↗

Friction removed Courts and tribunals Fair Work Commission, AU

FWC hit by wave of AI-assisted dismissal claims from Uber drivers

Unfair dismissal claims involving gig workers surged from virtually none to 500 in the first year, creating a backlog of unresolved cases at the commission.

Who pays: The Fair Work Commission tribunal through an administrative backlog and case delays

Scale Speed fast Adaptation none The Australian Financial Review ↗

Friction removed Science and publishing Academic computer science conferences (NeurIPS, ICLR, ICML, ACL)

Major AI conferences desk-reject papers over LLM-hallucinated citations

Four major machine learning and NLP conferences (NeurIPS, ICLR, ICML, ACL) instituted automated cross-checks of citations against bibliographic databases and desk-rejected submissions with non-existent references, causing over 100 desk rejections at ACL alone, elevated desk-rejection rates at ICLR, and emergency review processes for false positives.

Who pays: Program chairs, senior area chairs, and reviewers spending hours adjudicating flagged references, plus authors navigating erroneous desk-rejections caused by false positives

Scale Speed fast Adaptation verify StrictCite ↗

Friction removed Public consultation Parliament of Australia, AU

AI-generated submissions with fake references flood Australia's parliament inquiry system

At least 39 submissions to the current parliament contained fabricated references, over 100 carried ChatGPT metadata tags, and at least one committee report incorporated the hallucinated evidence.

Who pays: Parliamentary inquiries, lawmakers, and public policy decisions reliant on fabricated evidence

Scale Speed fast Adaptation none Complete AI Training ↗ 4News ↗

Friction removed Media and culture Instagram

Instagram Imposes Reach Penalties on Unlabeled and Mass-Produced AI Influencers

A surge in synthetic virtual influencers, including hundreds of AI-generated 'virtual doctors and health celebrities' promoting health products and giving medical advice to divert traffic.

Who pays: Instagram users and patients exposed to unverified medical advice and synthetic personas, and genuine creators competing against mass-generated accounts

Scale Speed fast Adaptation reform AIbase ↗

Signal collapse Courts and tribunals U.S. Federal and State Courts, US

Rise of Hallucinated Citations in Legal Filings Spawns Verification Tool Industry

Lawyers submit court briefs containing invented cases, altered holding details, and erroneous citations produced by AI editing tools, prompting judicial orders and sanctions.

Who pays: Judges, clerks, and opposing counsel who must cross-examine citations, along with law firms purchasing secondary verification software

Scale Speed fast Adaptation verify The Daily Record ↗

Signal collapse Hiring and work Greenhouse hiring platform, US

254 applicants per job: the AI doom loop in hiring

Applications look alike; recruiters cannot tell interest from automation; hiring slows.

Who pays: candidates who apply once and carefully

Scale Speed slow Adaptation price Fortune ↗

Arms race Science and publishing AAAI-26 conference

AAAI gave every paper an AI review, 22,977 in under a day

The reader automated; authors preferred the machine's review on technical accuracy.

Who pays: reviewers whose judgment is now benchmarked against a model

Scale Speed fast Adaptation detect arXiv ↗

Signal collapse Public consultation South Coast Air Quality Management District, US

More than 20,000 comments from one platform in an air-quality vote

A comment period read as overwhelming opposition; authenticity unresolved.

Who pays: residents whose real comments no longer count

Scale Speed slow Adaptation none Los Angeles Times ↗

Friction removed Open source and security curl bug bounty

curl ended its bug bounty after AI-written reports drowned the real ones

Confirmed-report rate from above 15% to below 5%; maintainers' time consumed by debunking; the reward removed.

Who pays: the maintainers, and legitimate reporters who lose the reward

Scale Speed fast Adaptation price daniel.haxx.se ↗

Arms race Science and publishing ICLR 2026 peer review

One in five ICLR 2026 reviews was fully AI-written

A fifth of reviews generated; authors answer filler; detection tools triage.

Who pays: authors, and the field's trust in its own review

Scale Speed fast Adaptation detect Pangram ↗

Friction removed Education California Community Colleges, US

AI-Assisted Ghost Students and Financial Aid Fraud at California Community Colleges

In 2024, California Community Colleges reported that across 116 campuses, nearly 31% of all applications received were fraudulent ghost students, causing $3 million in state funding losses and $10 million in federal funding losses.

Who pays: Public taxpayers funding federal and state financial aid, and college admissions and financial aid staff burdened with fraud verification

Scale Speed fast Adaptation verify Atkinson, Andelson, Loya, Ruud & Romo ↗

Friction removed Media and culture Amazon

AI Click Farms Flood Amazon Kindle Bestseller Rankings

Coordinated accounts released over 600 books within a single year, capturing roughly 86 of the top 100 spots in a single category and reaching the top 100 of the entire Amazon store.

Who pays: Human independent authors whose discoverability, chart placement, and Kindle Unlimited revenue are displaced by mass-produced AI books

Scale Speed fast Adaptation none Facebook ↗ · social source, review