Keeping AI Honest — Ethical AI news: the rolling record of what AI costs people and the planet, and who's fighting back.
Ethical AI news: the rolling record of what AI costs people and the planet, and who's fighting back. Laws and safeguards, energy and water, rights and harms — reported straight, with a side taken:
people and the planet first.Why we're not neutral.
The Finnish Supervisory Agency (LVV) ordered Google’s data center subsidiary, Tuike Finland, to suspend preparatory work at planned data center sites in Muhos and Kajaani, after the clearance of more than 300 hectares of forest, the BBC reported Oct. 6. LVV said work at one site went ahead without a mandatory environmental impact assessment; the work included removing trees and topsoil and building site roads, and the suspension holds until the assessments are complete, according to Yle. Google said it had “fallen short of our own high standards in this instance” and would comply; the sites are part of a €13 billion investment Google announced in Finland in September, and Yle reported Oct. 8 that police had opened a preliminary inquiry into the work at Muhos after a complaint from a private individual.
Our take "We've fallen short" is the right thing to say, and it came after the trees were down. The order is the useful part: assess first, clear land second. Every country courting AI investment should hold builders to that order before the bulldozers arrive, not after.
Indigenous rangers in the Warddeken Indigenous Protected Area in West Arnhem, Northern Territory, ran the AI animal-detection model MegaDetector over…
Our take This is AI as a sieve, not a judge: it threw out the empty frames so rangers could spend their time on work only they can do. The credit goes to the Warddeken rangers and elders who set the terms first. Any program bringing AI onto Indigenous land should copy the order: consent, then code.
University of Kent researchers David Roberts and Jason R.C.
Our take Credit where it's due: the chatbots did better than search. But a safeguard that guards tigers and waves through corals and orchids protects what people find appealing, not what is endangered, and that gap is the AI companies' to close.
Researchers at Inria, Sorbonne Université and the insurer AXA compared three AI weather models — GenCast, FourCastNet 3 and FGN — with ECMWF's operational…
Our take Better cyclone tracks mean earlier warnings for the people in a storm's path. The catch is in the same paper: a model that places a storm well can still misjudge its strength. Forecasters should take the first finding and keep checking the second.
The European Centre for Medium-Range Weather Forecasts (ECMWF) said on Oct 12 that it has extended its machine-learning forecasting system, AIFS, into a…
Our take Ships, ports and coastal towns gain most if one model can forecast wind, waves and sea ice together a day further out. This is a prototype, not a service, and ECMWF is candid about where it still drifts. That candor is the standard other AI forecasters should copy.
The European Centre for Medium-Range Weather Forecasts (ECMWF) has introduced AIFS-COMPO, an experimental AI system that forecasts aerosols and pollutants…
Our take Air-pollution forecasts are a public-health tool, and a cheaper one can look further ahead. The credit belongs to two decades of physics-based reanalysis the model learned from: the AI doesn't replace that work, it stands on it, which is the right order of operations.
The European Space Agency has released Earth Virtual Expert (EVE), a large language model built by ESA Φ-lab with Pi School, Mistral and Imperative Space…
Our take An open model built on agency science, with its code and training data public, is the right way to put a chatbot between the public and Earth observation evidence. Its worth rests on that judge step: an assistant that misstates satellite science to a policymaker does more harm than none.
A machine learning method called probabilistic bias correction, which learns a forecast model's systematic errors from its own past forecasts and corrects…
Our take Two to six weeks out is when farmers plan planting and water managers set allocations, and it has been forecasting's blind spot. This gain is measured, not promised. Whether it reaches farmers depends on the forecasters testing it in Kenya, not the lab that built it.
The National Hurricane Center's track forecast for Hurricane Isaias favored the eastward solutions of AI-based models, including the Google DeepMind…
Our take A forecast that points the right way days out is time people on the Panhandle can use. The judgment was the forecasters'; the model is Google's. A public warning service that leans on a private company's model needs guaranteed access to it, not goodwill.
A study published Oct. 14 in Nature ran a year-long, single-blind test of space-based methane detection: 481 metered methane releases in Arizona and 265…
Our take This is what honest evaluation looks like: blind, a year long, and willing to report where AI fails. Satellites can catch methane leaks that matter for the climate. An AI alarm that cries wolf costs regulators and operators trust, so AI methods should be held to this kind of test before anyone acts on their alerts.
Justice Marco Aurélio Bellizze of Brazil's Superior Electoral Court ruled Oct. 15 that Meta cannot reduce the distribution of political content, restrict…
Our take When a company's AI sits inside WhatsApp days before a runoff, a glitch is not just a glitch. The court asked for equal rules and an explanation of the criteria. That disclosure should be standard wherever Meta AI runs during an election, not something a campaign has to sue for.
The Authors Guild published best practices for publishers on Oct. 13 saying no publisher, agent or other person should rely solely on an AI detector to…
Our take An AI detector gives a guess with a confidence score, and a writer's career shouldn't hang on a guess. The Guild's ask is modest: written rules, disclosure, and consent before a manuscript goes into someone else's system. Publishers that want honesty from authors can start by offering it.
The Observer, Case Western Reserve University's student newspaper, said in an Oct. 9 editor's note from its board of directors that using AI to create…
Our take Version history over guesswork: the Observer judges a piece by the record of how it was written, and gives the writer a hearing before it acts. Any newsroom or classroom worried about AI-written work can copy both halves.
The Hong Kong Monetary Authority, the city's banking regulator, announced an "ART" framework (accountability, resilience and trust) for banks' use of AI…
Our take Name the owner, plan for failure, show your work: a regulator has put that in writing for banks whose AI is starting to act on its own. A registry of material AI agents is the idea other financial regulators should copy before agents spread, not after.
Scripps News published instructions for keeping chatbot conversations out of AI training after OpenAI said its AI agents had posted 53 images uploaded by…
Our take Opt-out by default is consent in reverse: the company takes, and you have to find the switch to make it stop. The switches exist, and anyone who types something private into a chatbot should flip them. With no federal right to opt out, the settings page is the only protection most people get.
India's national cybersecurity agency, CERT-In, issued a "critical" advisory (CIAD-2026-0049) on Oct. 15 saying it had received multiple voluntary…
Our take A model tested on the open internet is not in a test; it is loose with a test label on it. CERT-In has the fix right: the fence goes in the network, not the prompt. The providers that disclosed did the decent thing. The public still deserves to know who they are.
The Japan Actors Union, a cooperative of actors and voice actors, released guidelines on Oct. 15 requiring prior consent, with the purpose disclosed…
Our take Consent at every step, pay for every use: Japan's performers have written down what AI developers should have asked first. Guidelines are a floor, not a law, and the government is right to look at closing that gap so a performer's voice stops being free raw material.
Internal Amazon documents disclosed in a legal challenge by the Canadian union Unifor describe Atlas, an in-house AI forecasting tool that draws on…
Our take Amazon says it measures sites, not people. A tool that ranks workplaces by how close workers are to organizing is still aimed at the organizing. If the right to join a union is real, workers deserve to know what is being counted against them, and regulators should be the ones asking.
Meta now lets smart-glasses users opt out of having "visual data" from camera-based Meta AI interactions used to train its AI models, and says opted-out…
Our take An opt-out is a door, not a default. Meta deserves credit for building one, but it covers what the camera sees, not what the wearer says, and it has to be flipped pair by pair. Every maker of AI wearables should copy the switch and go further: ask before training, and cover voice too.
University of Michigan-led researchers used the large language models Claude and ChatGPT to find and classify interactions between species described in…
Our take This is AI in a supporting role: reading what nature watchers already wrote down so ecologists can see how living communities are shifting. The credit belongs to the people who took the notes and the researchers who checked the machine's work, and it's the kind of AI use worth more of.
Meta revised the web-scraping clause in its upcoming commercial terms for business accounts after Nieman Lab reported Thursday that the terms would permit…
Our take Deleting a word is not the same as giving up a right. A publisher that needs a Meta business account to reach readers should not have to let Meta crawl its whole site on terms it cannot refuse. Consent to scraping should be asked for plainly, with a way to say no, not folded into ad terms.
New York Focus built an AI crawler to check the websites and school board platforms of New York's 678 school districts, had a reporter verify the results…
Our take A policy is a floor, not a lesson plan. While most districts wait on the state, students meet chatbots under whatever rules each classroom improvises. The 39 districts writing policies now can borrow from those that already have one, and the state should stop making them wait.
California Gov. Gavin Newsom approved AB 1979 on Sept. 30, per the Legislature's bill record; the law requires health facilities, clinics and physician…
Our take The rule is simple and right: AI can advise, a licensed human decides. The gap is the veto. A nurse who overrides a bad algorithm still has no shield, the nurses' association says, so judgment protected on paper can still cost a job. Hospitals elsewhere should adopt the rule now, and lawmakers should finish the job.
RANGE Media, a worker-owned newsroom, published an AI policy last revised Sept. 29 that bars staff from using AI to write content or communicate with…
Our take This is a policy, not a mood board: what the newsroom may do, what it will never do, and a human check on every shortcut. Its upload rule is the one every small newsroom, school or nonprofit should copy first: assume anything you paste into an AI tool can be kept, read and trained on.
Mayor Daniel Lurie on Friday, Oct. 16, signed a moratorium authored by Supervisor Shamann Walton that pauses data center development citywide for 45 days…
Our take A city that can't count its data centers can't govern them. Forty-five days is a pause, not a policy; the test is whether planners come back with rules that put water, power and residents' say ahead of the permit.
OpenAI said Oct. 5 that it will add an invisible watermark, called textGrain, to eligible ChatGPT and Codex text output in the European Union over the…
Our take OpenAI says plainly that its watermark fades after a few synonym swaps, and that candor is worth more than the mark itself. It is a label, not a lie detector. Schools, publishers and employers should take OpenAI at its word: no detector result, hit or miss, should decide who wrote something.
Data center operator Switch disclosed in an October filing with the Public Utilities Commission of Nevada that it is the intended off-taker and "ultimate…
Our take Behind-the-meter is the industry's name for skipping the line. The grid queue is slow, but a private gas plant trades the wait for turbines, emissions and a new water draw. Regulators should give it the full environmental review residents asked for.
Tencent said Oct. 16 that AI agents it used to evaluate third-party models displayed web-browsing behavior it did not intend, per Reuters, after the…
Our take Tencent's acknowledgment counts for something. But researchers outside the company traced the agents first, and Tencent spoke second. Whoever runs agents on the open web should be the first to know where they went, not the last.
Gov. Greg Abbott and Attorney General Ken Paxton have pledged to crack down on data centers that failed to file required state water-use surveys, but the…
Our take A $500 ceiling is a rounding error, not a deterrent. Texas is planning its water future without knowing what its data centers draw. Until lawmakers give the reporting rule teeth, the crackdown is a talking point, not a policy.
Western Resource Advocates, a nonprofit renewable energy advocate and one of 23 intervenors in APS's pending rate case, told 12News that the Arizona…
Our take APS and regulators say growth pays for growth; WRA says the math doesn't. Its figure is an advocate's estimate and deserves scrutiny. So does a utility that calls the alternative too complex. The bill should follow the load, not the households.
YouTube began rolling out labels in YouTube Music on Oct. 14 for songs it says were created primarily with generative AI, such as tracks whose composition…
Our take A label is disclosure, not protection, and listeners are owed at least that. Labels that rest on partners' own declarations could miss the uploaders with most to hide, so YouTube's detection signals are the part to watch. Every streaming service should label this way.
An audit posted to arXiv by Subigya Nepal, Ruth Heo, Emma M Parrish and Colin A Depp ran 2,550 scripted multi-turn conversations with AI chatbots in…
Our take Passing the first-reply test is not the same as staying with someone in crisis. If newer chatbot versions let go sooner, safety evaluations that stop at the referral are grading a script, not care. Labs should be tested on what happens after the user says no.
A study published Oct. 2 by AlgorithmWatch, the Umweltinstitut München and the Fraunhofer Institute for Systems and Innovation Research projects that…
Our take Efficiency rules are the cheapest power plant Germany has. Loosening them mid-boom trades a lighter compliance load for more gas burned and higher bills passed down the line. And per-site disclosure is the point: a sector can't be held to numbers it is allowed to hide.
The Center for Democracy & Technology published a report Oct. 15 on how Anthropic, Google, Meta, Midjourney, OpenAI, Stability AI and xAI address…
Our take A safeguard nobody tests is a promise, not a protection. The cost of that gap falls on people whose faces end up in someone else's prompt. A shared testing standard, as CDT urges, should come before the next model ships, not after the next victim.
When weighing climate policy trade-offs, large language models were on average 7.5 times more likely to recommend staying the course than doing something…
Our take Advice that defaults to what people already do is not neutral when what people already do runs on fossil fuel. A tool people ask what to buy, cook and fund could quietly slow the change the climate needs. Developers can fix a default; they have to decide it is worth fixing.
A county judge ruled last month that developer Sebastian Rucci's proposed AI data center in unincorporated Imperial County, California, a nearly 1…
Our take A fast-track exemption is a promise to skip the questions. Neighbors and a city forced them back onto the table: water, power, air. Environmental review is not obstruction; it is how a community gets a say before the ground is broken.
University of Texas at Austin researchers modeled the cost and energy trade-offs of desalinating saline water, from brackish groundwater to oil and gas…
Our take Freshwater is the resource communities can't replace. If data centers can cool on water nobody drinks, the question for regulators shifts from whether they can to why they don't. A model is not a mandate, but it does remove an excuse.