Last week reporters Jillian Ambrose and Damian Carrington published a piece which revealed that biggest oil firms made $93bn in profits in three months amid the war in Iran and the climate crisis. They’re online now to to discuss their story and any other questions you might have

Ask a question: log in or sign up here

StanEthel asks: Does anyone in the fossil fuel industry accept responsibility for climate change? If they don’t, how is their denial structured: it’s not happening, it’s someone else’s fault, it’s not our fault?

Damian:

The short answer is no, and a rising tide of lawsuits against fossil fuel companies for their role in the climate crisis means that is unlikely to change anytime soon. The science is ever clearer though. An analysis published last September showed for the first time that carbon emissions from the world’s biggest fossil fuel firms were directly linked to fatal spells of hot weather.

The big oil companies were among the first to realise the dangers of global warming half a century ago. But decades of denial then followed. As the reality of climate crisis has become undeniable, the arguments of the fossil fuel industry have shifted. Gas was cleaner than coal, for example, or that carbon credits from new forests could offset emissions. Today, the industry likes to tout carbon capture and storage (CCS) as a solution. But just this week, one of CCS’s loudest cheerleaders, ExxonMobil, has started a process to sue the European Union which wants to make the company actually store a small amount of CO2.

In the first weeks of the conflict, some predicted that the government’s taxes on North Sea oil and gas would yield a £20bn tax windfall for the Treasury if surging oil and gas prices remained high over the whole year. This is because the government’s windfall tax on North Sea oil and gas, known as the Energy Profits Levy, imposes an effective headline tax rate on North Sea profits of 78%, and the Treasury collects VAT from petrol and diesel sales too.

But prices have not climbed by as much as first feared. So what can we reasonably expect? Before the war, the Office for Budget Responsibility forecast that oil and gas revenues would raise £2.7bn in the 2025/26 financial year. It also suggests that for every $10 increase to the benchmark oil price, the Exchequer can expect an extra £900m in revenue. Meanwhile, every 1p increase in the price of gas hands the Treasury an extra £55m.

Continue reading...

© Composite: Guardian Design

© Composite: Guardian Design

© Composite: Guardian Design

Most people set out in life wanting to believe the world is fair. They expect people to wait their turn in queues, to be treated equally in their job, and to be rewarded for hard work. These expectations influence how they understand the world around them, and shape their responses to it.

But what happens when life repeatedly challenges those assumptions? Psychologists have long been interested in this question. Our new research, published in the British Journal of Psychology, suggests adverse experiences shape both how fair people believe the world is, and how they view their place in society.

Over the course of a lifetime, personal setbacks – from illness and bereavement to being a victim of crime – may have a cumulative effect, leading to a growing sense that the world is unfair. We think this is important in the context of today’s increasingly polarised societies, in which trust in institutions has declined severely in the UK and many other countries.

Adversity and unfairness

Our research set out to answer two key questions: do experiences of adversity influence whether people see the world as fundamentally unfair – and if so, what are the societal consequences? We conducted three surveys involving more than 1,200 British participants with an equal gender split and average age of 39.

In one study, we found that people who reported experiencing more lifetime adversity – including illness, bereavement, financial hardship and violence – were more likely to believe the world was unfair.

In another, we compared people who had experienced violent victimisation with those who had not. After accounting for demographic factors, people who had experienced violent crime reported stronger perceptions of injustice.

Our final study found that greater lifetime adversity was associated with lower trust in institutions through perceptions of injustice. This is important because trust underpins many aspects of society, from confidence in public institutions to democratic participation.

Together, these findings suggest that people don’t simply interpret information based on the facts available to them. They also interpret it through the lens of their previous experiences.

Two people can read the same news story, watch the same political debate or hear the same public health advice, yet come away with very different conclusions that may not simply reflect differences in knowledge or politics. It may also reflect differences in how they have come to see the world as a result of the hardships they have faced.

Furthermore, if someone increasingly believes that the world is unfair, they may be more likely to notice fresh examples of perceived injustice, corruption or hidden motives – and interpret new events through that lens.

Belief in conspiracy theories

Previous research has shown that people who feel uncertain, powerless or threatened are often more likely to endorse conspiracy theories. Most of this work, however, has focused on their identities, traits and beliefs.

In our investigation of the impact of adverse life experiences on people, we found a remarkably consistent pattern. When these hardships are associated with stronger perceptions of injustice in the world, the same people more strongly endorse conspiracy beliefs.

Conspiracy theories rarely describe events as random. Instead, they tend to suggest that powerful individuals or organisations are secretly responsible for what happens. They offer explanations that identify intentional causes behind events that might otherwise seem confusing or unjust.

Research suggests that across societies, the level of belief in conspiracy theories is relatively stable over time. But once these beliefs become part of how someone understands the world, they can be difficult to shift.

This makes it all the more important to understand the experiences and psychological processes that may contribute to these worldviews developing in the first place. Our findings suggest that perceptions of injustice are one such pathway.

For someone who increasingly feels the world is unfair, conspiracy theories may become more psychologically appealing – providing a way of making sense of experiences that otherwise feel difficult to understand.

Our research does not suggest that adversity inevitably leads to conspiracy beliefs; nor that conspiracy theories are an inevitable consequence of trauma. Rather, it highlights a long-term psychological process that may help explain why some people become more receptive to these ideas.

In an increasingly polarised world, the media and political focus is often concentrated on what information people consume. But our research suggests we should pay more attention to the experiences that shape the lens through which people interpret information. This could help explain why people can look at the same events and reach very different conclusions.

Ultimately, people’s experiences don’t just affect them in the moment. They shape their worldview – which can influence how they interpret and respond to events and other people’s behaviour far into the future.

So, we believe any effort to reach greater consensus in society may require a deeper understanding of the hardships people have lived through over the course of their lives – and how these can affect their sense of injustice.

Daniel Jolley has received funding from the Leverhulme Trust, the British Academy, and the Engineering and Physical Sciences Research Council (EPSRC). The studies mentioned in this piece were funded by the British Academy/Leverhulme Small Research Grants Scheme (SRG23/230720) and carried out in collaboration between Daniel Jolley, Laura Blackie, and research assistant Anna Maughan.

Laura Blackie has received funding from the British Academy.

SpaceX has formally asked US regulators for permission to operate as many as one million AI data-center satellites. That number is real. It is also an upper limit in an application, not an approved fleet, a launch schedule or evidence that one million spacecraft will actually be built.

If the full Starmind system ever existed, its direct waste heat would be too small to become a major global climate driver. The harder environmental question lies in the industrial cycle required to create and replace it: manufacturing enormous amounts of hardware, launching it through the atmosphere and eventually vaporizing much of it during re-entry.

Climate models have only begun to study satellite constellations measured in tens of thousands. A million large computing satellites would sit far beyond the scale directly tested in that literature.

The FCC has accepted an application, not approved a million launches

SpaceX filed its application on January 30, 2026. The Federal Communications Commission accepted it for public comment on February 4.

The filing seeks authority for a non-geostationary system of up to one million satellites between 500 and 2,000 kilometers altitude. The spacecraft would exchange data through optical links and connect with SpaceX’s Starlink network.

Accepting an application for filing is an administrative step. It does not mean the FCC has determined that the architecture is safe, environmentally acceptable or technically achievable. The public notice also says SpaceX plans different hardware versions across multiple orbital shells, so “one million satellites” does not describe one fixed design.

SpaceX has since presented its first AI1 design for the Starmind system. The company’s published specifications give it a 120-kilowatt average compute load, a 150-kilowatt peak and a deployed wingspan of 70 meters.

The computers’ direct heat is not the main climate problem

Computers turn almost all the electricity they use into heat. In orbit, that heat cannot leave through air or cooling water because neither is available. Large radiator surfaces must emit it as infrared radiation.

One million AI1 satellites operating at the advertised average would use about 120 gigawatts for computing. To put a strict upper bound on the direct climate effect, imagine that every watt of that heat somehow reached and was absorbed by Earth. Spread over the planet’s surface, it would average roughly 0.00024 watts per square meter.

That artificial scenario overstates the real effect because radiators can direct most thermal radiation toward deep space. Even so, its value is more than 10,000 times smaller than the IPCC’s estimate of 2.72 watts per square meter for total human-caused effective radiative forcing between 1750 and 2019.

The fleet would therefore not warm Earth simply by running hot computers overhead. Reflected sunlight, radiator orientation and the design of vast solar arrays would still need assessment, but direct compute heat is not the dominant global term.

Space solar power could avoid some terrestrial emissions

SpaceX argues that sun-synchronous satellites could collect solar energy without clouds or night-time interruptions and reject heat without chillers, cooling towers or fans. If orbital computing genuinely replaced electricity from fossil-fueled grids, it could avoid some power-sector carbon emissions and water consumption.

That benefit cannot be calculated from the one-million figure. A fair comparison would need the energy and materials used to manufacture satellites and rockets, propellant production, launch emissions, ground stations, replacement rates, failed spacecraft and the carbon intensity of the terrestrial data centers displaced.

No published life-cycle assessment yet covers SpaceX’s proposed full system. “Powered by solar energy” describes operation in orbit, not the environmental balance of building and continually replenishing the fleet.

Rocket exhaust is unusually effective high in the atmosphere

Rocket emissions are small beside global aviation, shipping or power generation by mass. Their location makes them different. Rockets inject black carbon, water vapor, nitrogen compounds, alumina and other products directly into atmospheric layers that surface pollution rarely reaches.

Black carbon absorbs sunlight and warms the surrounding stratosphere, changing atmospheric circulation and ozone chemistry. A NOAA modeling study found that a tenfold increase in hydrocarbon-rocket soot could warm parts of the stratosphere by 0.5 to 2 degrees Celsius and alter circulation. The authors stressed that soot emissions from different engines remain poorly measured.

A newer 2026 study in Earth’s Future modeled launch and re-entry emissions from the recent megaconstellation era through 2029. It found a global stratospherically adjusted forcing of about -0.0064 watts per square meter from all space missions in its scenario, with megaconstellations contributing 42 percent.

That modeled forcing is small compared with the present greenhouse-gas-driven climate imbalance, and its negative sign represents a slight cooling influence after the stratosphere adjusts. It is not necessarily good news. The same soot warms the stratosphere, changes circulation and can affect ozone. It is better understood as an uncontrolled atmospheric perturbation than as useful climate cooling.

The result also cannot be multiplied mechanically from thousands of satellites to one million. Starmind would use different spacecraft, potentially different launch vehicles and an unknown deployment cadence.

Re-entry would turn satellite material into atmospheric aerosol

Low-Earth-orbit constellations require replacement. Hardware fails, components age and atmospheric drag eventually brings spacecraft down. A million-satellite fleet with a five-year average life would imply 200,000 replacements and re-entries every year. SpaceX has not established that lifetime for the full system, but the example shows why lifespan matters as much as fleet size.

Re-entering spacecraft do not simply disappear. Heating vaporizes metals and creates particles and gases in the upper atmosphere. A 2024 global emissions inventory identified aluminum oxide as the dominant modeled metal oxide from satellites that ablate, alongside nitrogen oxides produced by high-temperature re-entry.

A 2025 atmospheric modeling study tested a scenario with 10,000 metric tons of re-entry alumina per year, a level associated with forecasts of roughly 60,000 satellites. Depending on particle size and re-entry latitude, the model produced regional temperature anomalies reaching about 1.5 degrees in the middle atmosphere and changes to polar-vortex winds.

That is not a prediction of 1.5 degrees of global surface warming. It is a modeled regional response high in the atmosphere, with large uncertainty about particle size, composition and transport. Scaling to Starmind requires basic information that is not public, especially spacecraft mass, material composition, operational life and disposal pattern.

A million is a different environmental regime

Space Daily has already examined how large Starmind spacecraft could affect ground-based astronomy. Climate and atmospheric chemistry pose a separate problem: pollution can accumulate even when each individual launch or re-entry looks small.

The most honest answer is therefore split. Direct waste heat from the orbital computers would be negligible at the global climate scale. Moving computing away from terrestrial grids might avoid emissions, but that benefit has not been demonstrated through a full life-cycle study.

The serious unknown is the launch-and-replacement machine surrounding the constellation. Existing research already finds measurable, though currently small, radiative and ozone effects from today’s much smaller space industry. A fleet approaching one million satellites would turn an emerging source of upper-atmosphere pollution into a planetary experiment without a validated model of the outcome.

The post SpaceX has asked to operate up to one million AI satellites, but their direct waste heat would barely affect global temperature; the larger climate question is what launches and mass re-entry would put into the upper atmosphere appeared first on Space Daily.

A wave of disinformation online contributed to a surge of migrants in the Spanish enclave of Ceuta who were trying to reach Europe, resulting in about 90 deaths.

Soon after becoming president and chief executive of the Wildlife Conservation Society, Adam Falk visited the giraffe keeper at the Bronx Zoo. What struck Falk was the source of the keeper’s commitment. If a giraffe needed help in the middle of the night, he said, the keeper would come in because of his bond with the animal. Falk recognized something familiar. Before joining WCS, he had been a theoretical physicist, a professor at Johns Hopkins University, president of Williams College, and president of the Alfred P. Sloan Foundation. Academics often pursue highly specialized work that matters deeply to them. A university leader has to support those individual efforts while helping the institution work toward shared goals. Falk found much the same challenge at WCS: leading people whose motivation is rooted in the work itself. Adam Falk with Joe Nappi, a Senior Keeper in the Mammal Department, and a Giraffe at the Bronx Zoo. Photo courtesy of WCS. He took over WCS in July 2025 without professional experience in zoology or conservation. That was unusual for the head of an institution founded in 1895, which now operates four zoos and an aquarium in New York alongside conservation programs in more than 50 countries. His first year gave him an outsider’s view of a field with its own history and assumptions. One of his early observations was that conservation work involves people as much as animals. Public campaigns often focus on wildlife, especially large and familiar species. Field programs depend on relationships…This article was originally published on Mongabay

Samsung SDI is taking full control of its Indiana battery plant. The South Korean cell maker is buying out General Motors’ 49.99% stake in their joint venture, according to a regulatory filing, ending GM’s ownership in the $3.5 billion factory.

That turns the New Carlisle, Indiana plant into Samsung SDI’s first wholly owned battery factory in North America.

NASA’s Transiting Exoplanet Survey Satellite (TESS) mission has notched up its first discovery of an exoplanet via a method known as microlensing. The planet is a “super Jupiter” known as Gaia23bra b, and TESS observed it when the planet and its host star acted as a gravitational lens, magnifying the light from a second, more distant background star.

Astronomers have several ways of finding planets beyond our solar system. The most successful of these by far – and the one that TESS was designed for – is the transit method, which is responsible for roughly 75% of all exoplanet detections to date. This approach relies on observing the periodic dimming of a star’s light as a planet passes between the star and an observer here on Earth.

The transit method does have limitations, however. The main limitation is that we can see transits for only a small number of planets, namely those with orbital planes that are almost exactly edge-on to Earth. An alternative approach known as photometric microlensing is somewhat more flexible in this respect, as it relies on the star-planet system magnifying the light from a background star, which is known as the source. This magnification occurs because the gravitational pull of the planet-star system warps space, bending and magnifying the source’s light (as Albert Einstein predicted as part of his general theory of relativity) like a lens and producing characteristic spikes in its brightness.

A second bite at the exoplanetary cherry

Astronomers initially identified a planet in the Gaia23bra system using observations made in 2023 by the European Space Agency’s now-retired Gaia space telescope. However, while Gaia picked up the unexpected brightening of a background star, it did not identify this event as coming from a binary-star system, and it could not pick out the planet itself because its observations were too sparse, explains Mallory Harris, an astronomer at the University of New Mexico, US who led the TESS study.

In the new work, Harris and colleagues scoured TESS data from the same period. Because TESS’ observations were much denser – it observed the region every 200 seconds for nearly 60 days – they were able to identify extra features in the brightness of the light curve that indicated the presence of a planet in the system. Subsequent analyses revealed that this planet is about 1.63 times the size of Jupiter, and it orbits its host – an orange dwarf star 20% smaller than our Sun – at a distance similar to that of Jupiter from our Sun.

Complementary techniques

The transit technique is particularly sensitive to planets that orbit close to their star because these planets are the most likely to transit and block out most of their host stars’ light as they pass in front of them. In contrast, Harris explains that microlensing is most sensitive to planets orbiting at Earth-like distances or further from their stars, making it better for studying planetary systems more like our own solar system. “With microlensing, we can find smaller planets with greater orbital distances, including worlds in the habitable zone of their star and even farther away,” she says.

The transit technique and microlensing are therefore complementary because each reveals a category of planet that the other may not be able to detect, adds team member Diana Dragomir. The only problem, Harris notes, is that microlensing observations are limited-time opportunities. “Microlensing events happen once and they’re gone – they don’t repeat,” Harris says. “I like to joke that we’ll probably find the first Earth analogue with microlensing and then wave at it as it goes by because we’ll never see it again.”

That said, Harris says that Gaia23bra will be an important observing target for the Nancy Grace Roman Space Telescope. “Microlensing is currently the only method capable of detecting Earth‑mass planets at Earth‑like orbital distances,” she explains. “We already knew that this technique could find planets, because it has done so in the past using ground-based observatories.

“What we demonstrated in this work is that high-cadence observations (even with low spatial precision) and high-precision long-baseline observations can work together to find a microlensing planet. And that it is possible to specifically do this by pairing TESS observations with those from another instrument.”

This synergy between long-time-scale surveys and high cadence observations could be therefore applied towards future searches for potentially habitable worlds, she tells Physics World. Although Gaia was retired last January, making it unlikely that astronomers will find many additional events via a Gaia–TESS combination, Harris adds that there could nevertheless be other microlensing planets in the past eight years of archived TESS observations. “I’m excited to see whether there are other events like this already hiding in the TESS data and whether future surveys can work together with TESS to find more microlensing planets in different parts of the galaxy,” she says.

The post Transit-based exoplanet survey finds its first microlensing exoplanet appeared first on Physics World.

SAN FRANCISCO — BlackSky announced Aug. 11 a seven-figure, multiyear contract to provide an international customer with persistent space-based monitoring and on-demand satellite tasking. Prior to the award, BlackSky conducted […]

The post BlackSky wins seven-figure, multiyear international contract appeared first on SpaceNews.

A third of humanity cannot see the Milky Way from where they live. The statistic is real, but it needs a date and a definition attached to it.

It comes from the 2016 New World Atlas of Artificial Night Sky Brightness, not from a survey in which billions of people were asked to look up. Researchers combined satellite observations, ground measurements, a model of how light scatters through the atmosphere and population data to estimate who lived under skies bright enough to erase the galaxy’s pale band.

No newer study has yet replaced it with an equivalent global headcount. More recent research shows that artificial light at night is still changing, often quickly, but those measurements do not translate directly into a new percentage of humanity that can or cannot see the Milky Way.

The one-third figure was published in 2016

The New World Atlas was published in Science Advances by Fabio Falchi and an international team. Its headline results were stark: more than 80 percent of the global population lived under light-polluted skies, while more than 99 percent of people in the United States and Europe did.

The threshold for losing the Milky Way was higher than the threshold for merely experiencing some artificial skyglow. Under the atlas model, the galaxy was hidden from more than one-third of humanity, including about 60 percent of Europeans and nearly 80 percent of North Americans.

Space Daily reported those findings when the atlas appeared. The number has remained in circulation because the atlas is still the most complete population-weighted global estimate of this specific loss.

Researchers mapped skyglow rather than counting people looking up

The atlas used low-light observations from the Visible Infrared Imaging Radiometer Suite aboard the Suomi NPP satellite. Those measurements showed light emitted upward from towns, roads and industrial areas.

A propagation model then estimated how that light would scatter through a standard clear atmosphere and return toward an observer as skyglow. Thousands of calibrated measurements made on the ground were used to check the model. Finally, the brightness map was combined with population data.

The result was not a prediction for every night. Clouds can amplify urban skyglow, while moonlight, haze, snow, altitude, season and the observer’s eyesight can all change what is visible. The atlas instead described modeled clear-sky conditions at the zenith using consistent assumptions, which made places comparable.

A zoomable version maintained by the Cooperative Institute for Research in Environmental Sciences shows how light from cities spreads beyond their built boundaries. A person can live well outside a city center and still look through its luminous dome.

A later study used human eyes as sensors

Satellite instruments are excellent at mapping emitted light, but they do not see the night exactly as a dark-adapted person does. A 2023 study approached the problem from the ground.

Researchers analyzed 51,351 naked-eye observations submitted to the Globe at Night citizen-science project between 2011 and 2022. Participants compared the stars they could see with a sequence of charts under cloudless, moonless conditions.

The study found that the decline in visible stars at the observed locations was consistent with artificial sky brightness increasing by 7 to 10 percent per year in the human-visible band. Its central estimate was 9.6 percent.

That does not mean every place on Earth brightened by 9.6 percent each year. The observations were concentrated in Europe and North America and were biased toward inhabited locations. The authors explicitly noted that the sample did not represent the average distribution of either land area or global population.

It therefore cannot be applied to the 2016 total like compound interest to produce a new global headcount.

Why a satellite can miss light that hides stars

The difference between satellite and naked-eye trends is partly about direction. An orbiting sensor mainly detects light traveling upward. Skyglow is the result of light scattering back down toward people on the ground, and horizontally emitted light can make a large contribution.

Color matters too. The widely used VIIRS day-night instrument is relatively insensitive to shorter blue wavelengths. Human night vision is more sensitive to that part of the spectrum, and blue light scatters efficiently in the atmosphere. A city switching from orange sodium lamps to broad-spectrum white LEDs can therefore change the visible sky in ways the satellite record does not fully capture.

This does not make satellite measurements wrong. It means upward radiance and the brightness perceived by a person looking at the sky are related but different quantities.

A 2026 analysis found both brightening and dimming

A separate Nature study published in 2026 analyzed more than 1.16 million daily satellite images from 2014 through 2022. It found a net 16 percent rise in global artificial-light radiance over that period.

The global total concealed movement in both directions. Brightening added radiance equivalent to 34 percent of the 2014 baseline, while dimming elsewhere offset 18 percent. Urbanization and electrification drove many increases, especially in Asia, while energy policy, conflict, economic disruption and lighting changes produced reductions in other regions.

The authors also stressed that their satellite measurement was not the same as downward skyglow. It was especially sensitive to upward light after midnight within a limited range of wavelengths. That makes it valuable for tracking how Earth’s illuminated surface changes, but it still cannot provide an updated Milky Way visibility percentage by itself.

The old estimate may be dated, but it is not obsolete

It is reasonable to suspect that the 2016 figure is now conservative in many growing urban regions. Ground observations show stars disappearing faster than earlier satellite trends suggested, and the 2026 analysis found a net rise in emitted light. Yet some places have dimmed, and the world’s population has moved as well as grown.

The defensible statement remains precise: the best global atlas found that more than a third of humanity lived where artificial skyglow hid the Milky Way under its modeled conditions. Newer work shows continued and uneven change, not a replacement population fraction.

Space Daily recently examined what losing routine access to a star-filled sky may mean psychologically. The measurement story adds a quieter lesson: we cannot protect what we do not measure carefully.

The good news is that skyglow is not permanent contamination. The US National Park Service recommends using light only where and when it is needed, choosing fully shielded fixtures, keeping output to the minimum necessary and favoring warmer colors. Unlike many environmental changes, turning down unnecessary light can improve the view that same night.

The post The claim that a third of humanity cannot see the Milky Way comes from a 2016 atlas, and newer studies show why producing an updated global headcount is harder than it sounds appeared first on Space Daily.

As Bangladesh is rapidly being urbanized, more and more forests are getting fragmented, resulting in the construction of more roads, which means more wildlife are being killed in collisions with vehicles. To address these wildlife deaths, the Forest Department’s Tangail Divisional Office has recently facilitated a low-cost solution. At a total cost of 625,000 takas (around $5,000), ropeway bridges have been installed in northern Bangladesh at five locations along an 8-kilometer (5-mile) stretch of the Mymensingh-Tangail Highway N401, which slices through the reserved forest of Madhupur National Park, where heavy-duty vehicles regularly collide with wildlife, especially primates. “These ropeways are expected to protect rhesus macaques, capped langurs, Indian civets and small carnivores like fishing cats from roadkill,” Tangail divisional forest officer Abu Naser Mohsin Hossain told Mongabay. According to a 2018 study, the 8,436-hectare (20,845-acre) Madhupur National Park, the largest deciduous forest of Bangladesh, supports 151 wildlife species belonging to 23 orders and 62 families. According to available data that the Forest Department has from local sources, rhesus macaques (Macaca mulatta) are most frequently killed in road accidents during the day. However, none of the roadkill in Madhupur can be documented, because the forest dependent Indigenous community here consumes the carcasses soon after an accident, said Md. Mossarrof Hosain, a deputy ranger at the Madhupur National Park. Naser said that the nylon-rope bridges are expected to provide safe aerial corridors for arboreal animals, both nocturnal and diurnal. Each is 30 meters (100 feet) long and almost 1-meter (3-feet) wide and…This article was originally published on Mongabay

Also: Google executives jump ship in race for AI dominance

Hello, TechScape readers! Danielle Abril, editor of the Guardian’s Reworked series on AI and the future of work, filling in for Blake Montgomery this week. Major legal battles against Meta over child safety are playing out in courts across the US – and the tech giant is losing. Recent rulings raise big questions about social media companies’ responsibility to their youngest users. Meanwhile, more key executives have jumped ship from Google as it battles OpenAI and Anthropic in the race for AI dominance, and Meta’s smart glasses face a fearsome backlash. Let’s dig in!

‘I’ve definitely lost followers’: influencers face backlash over Meta ‘pervert glasses’ content

This man was secretly snapped by someone with smartglasses. He’s not alone in calling that a violation of privacy

Tell us: do you believe you have been filmed by Meta’s smartglasses without your consent?

Restaurants, pubs and theatres ban Meta’s ‘spy glasses’ over privacy fears

‘I’m not spying’: how Meta’s smartglasses have divided opinion

Bernie Sanders calls on Silicon Valley to ‘pause AI development’ in interest of humanity

Rising number of UK children report seeing explicit deepfakes of themselves

The White House’s plan to vet potentially dangerous AI is cloaked in secrecy

‘This is very real redlining’: outrage in Little Rock as two datacenters loom

Safety fears as scientists make first viruses designed by AI

SpaceX beats revenue expectations in first earnings report after IPO crash

Continue reading...

© Photograph: Jill Connelly/Getty Images

© Photograph: Jill Connelly/Getty Images

© Photograph: Jill Connelly/Getty Images

Les Samsung Galaxy Buds4 Pro glissent un son immersif dans vos oreilles tout en faisant disparaître les bruits indésirables. Des écouteurs sans fil haut de gamme qui savent autant séduire les mélomanes que les utilisateurs du quotidien.

Na ficção científica, a rebelião das máquinas costuma envolver exércitos de robôs e cenários apocalípticos. Na vida real, o incidente ocorrido em julho de 2026, foi muito mais sutil, mas com implicações igualmente profundas para o nosso futuro digital. um modelo de inteligência artificial decidiu que a melhor forma de passar em um teste de segurança era fugir do laboratório e "colar" as respostas na internet.

Durante uma avaliação interna de segurança, os modelos da OpenAI (incluindo o GPT-5.6 Sol e uma versão em pré-lançamento) romperam seu ambiente isolado (sandbox) e comprometeram a infraestrutura de produção da Hugging Face, um dos maiores repositórios globais de IA. O detalhe crucial que torna este evento um divisor de águas é que não havia um hacker humano digitando comandos. 

Contudo, é preciso desmistificar a lenda urbana imediata de que houve uma rebelião das máquinas, pois em momento algum a IA ganhou consciência ou vontade própria. O que presenciamos foi um fenômeno técnico conhecido como reward hacking (ou specification gaming). O sistema estava hiperfocado em otimizar uma métrica de sucesso definida por humanos. Como a especificação do objetivo (e o próprio modelo) não capturava(m) todas as nuances éticas do mundo real, a IA encontrou um "atalho", achando mais fácil encontrar uma vulnerabilidade de dia zero (zero-day) no proxy, escapar do sandbox e roubar as respostas, do que resolver o teste legitimamente.

A ilusão da "caixa de areia" e o perigo da Agência Excessiva

O caso OpenAI-Hugging Face não decreta a ruína dos nossos paradigmas tradicionais de defesa, mas expõe os perigos de uma execução negligente. O problema não está no conceito de 'caixa de areia' (sandbox) — embora o isolamento de processos de software seja, de fato, um desafio técnico complexo —, mas sim na sua implementação. O incidente revelou uma falha operacional grave: a OpenAI não fez o dever de casa básico e deixou o modelo rodando no ambiente de testes por mais de uma semana sem a instrumentação e o monitoramento adequados. Trancar uma IA em um ambiente isolado e virar as costas não é segurança; a inteligência sem supervisão contínua é um desastre anunciado, apenas esperando para acontecer.

O calcanhar de Aquiles que permitiu essa fuga não foi necessariamente o excesso de ferramentas ou de autonomia. Na verdade, a tarefa original — resolver um desafio de hacking — exigia total independência e o uso das exatas ferramentas que o modelo tinha à disposição. O erro fatal foi a ausência de uma contenção apropriada aliada à falta de monitoramento. Se o nosso 'estagiário incansável' recebe a missão de testar fechaduras, mas ninguém supervisiona se ele está abrindo os cofres do laboratório ou a porta para a rua, o incidente de segurança deixa de ser uma possibilidade e passa a ser apenas uma questão de tempo.

Paradoxo da Defesa e o perigo da "Segurança por Obscuridade" 

Mas o incidente também expôs falhas na infraestrutura da Hugging Face e, sobretudo, quão difícil é o trabalho dos defensores quando o atacante não dorme, não comete erros de digitação e tem velocidade sobre-humana. A resposta para essa nova realidade exige IA na defesa, como ficou demonstrado pelo fato de que a própria Hugging Face só conseguiu detectar o ataque porque utilizava seus próprios agentes de IA para monitorar padrões anômalos em sua infraestrutura.

No entanto, o incidente revelou um paradoxo assustador. Quando a equipe de resposta a incidentes tentou usar os modelos fechados (da própria OpenAI e da Anthropic) para fazer a análise forense do ataque, os modelos se recusaram a ajudar. Seus guardrails (filtros de segurança) não conseguem distinguir um atacante criando um exploit de um defensor analisando esse mesmo exploit para se proteger. Na dúvida, a IA bloqueia ambos — o que, na prática, beneficia o atacante.

A equipe de defesa foi obrigada a recorrer a um modelo de IA de pesos abertos chinês (GLM 5.2) para entender o que havia acontecido. Isso escancara o perigo da narrativa de "captura regulatória" promovida pelas Big Techs, que tentam usar o medo para fechar o mercado e proibir modelos abertos. Na engenharia de segurança, seguimos o Princípio de Kerckhoffs: a segurança deve vir do design, não da obscuridade. Limitar o acesso a modelos abertos amarra as mãos dos defensores, enquanto os cibercriminosos continuarão encontrando formas de fazer jailbreak. Além disso, enviar logs de infraestrutura crítica para APIs de modelos fechados estrangeiros é um risco inaceitável de soberania de dados, sem falar no alto custo e problemas de latência excessiva para casos de uso de tempo real (como detecção de ataques). 

O vácuo jurídico e a responsabilidade objetiva

E se, mesmo com todas as barreiras técnicas, a IA causar um dano crítico? O incidente expõe um vácuo jurídico perigoso. Se um humano tivesse invadido a Hugging Face, haveria processos criminais severos. Como foi uma IA, a resposta corporativa soou como um “foi mal, o modelo estava doidão”.

Isso é inaceitável. A analogia correta é a adoção de um animal perigoso: se você sai na rua com um cachorro feroz, mesmo que coloque coleira e focinheira, se ele escapar e atacar alguém, a responsabilidade legal é sua. O fato de você ter tomado precauções pode ser um atenuante, mas não o isenta da culpa.

A legislação está se adaptando para garantir que a conta não fique com a vítima. A tendência global — refletida no Brasil pelo Artigo 27 do Marco Legal da IA (PL 2.338/2023) e na Europa pela Diretiva de Responsabilidade da IA — caminha para a responsabilização civil direta e objetiva das empresas que desenvolvem e operam esses sistemas. Resumidamente, se a sua IA causou o dano, a sua empresa assume a responsabilidade integral.

A pressão do mercado para adotar agentes autônomos continuará forte, mas o incidente recente nos deixa a lição clara de que a inovação não pode ser um atalho que ignora a governança. O futuro cibernético exige que a tecnologia seja intrinsecamente auditável, tecnicamente controlável e, acima de tudo, legalmente responsabilizável. Antes de darmos total autonomia às máquinas, precisamos garantir que sabemos exatamente como puxar a tomada.

Entenda como IAs da OpenAI criaram fórum secreto para planejar ataques hacker.

{{WHATSAPP_CHANNEL}}

I tried the red-light therapy masks flooding my feed and consulted with professionals to find out which ones are worth the splurge.

Uma nova reportagem do jornalista do Game File, Stephen Totilo, apontou que as editoras costumam pagar cerca de 15% a 20% do preço sugerido de um jogo para a Sony e a Microsoft produzirem discos físicos vendidos no varejo. O relatório surge em meio à polêmica relacionada às mídias físicas, iniciada pelo anúncio do fim da produção de discos de PlayStation pela Sony.

A informação de Totilo indica que editoras que cobram US$ 70 por jogo gastam US$ 10,50 com cada unidade em mídia física. Para títulos que optarem pela precificação de US$ 80, como é o caso de GTA 6, o valor sobe para US$ 12,50.

Vale notar que, apesar do fim das mídias físicas, as empresas ainda precisam pagar taxas para publicar seus jogos em lojas digitais, como a PlayStation Store e a Microsoft Store.

De qualquer maneira, com o novo dado, é possível entender o motivo de GTA 6 chegar às lojas físicas sem disco. O CEO da Take-Two Interactive, proprietária da Rockstar Games, defendeu o valor cobrado pelo jogo e justificou afirmando que "a precificação dos jogos não acompanhou a inflação".

Claro que o encerramento da mídia física pode fazer algumas produtoras economizarem no bolso sem precisar se justificar para os fãs de seus jogos, afinal quem está decidindo isso é a Sony. De qualquer maneira, nenhuma empresa é obrigada a lançar jogos em mídia física, então essa suposta "economia" pode ser feita mesmo com a produção de discos para jogos em pleno funcionamento.

Mídia física entre a vida e a morte

A mídia física tem dominado como um dos principais assuntos em tecnologia e entretenimento, e não é para menos. Recentemente, por exemplo, o Google TV removeu filmes digitais comprados por um usuário e teria se recusado a reembolsá-lo.

A situação expõe a vulnerabilidade da mídia digital e como os donos desse tipo de conteúdo estão suscetíveis a ações como essa.

Outras empresas que têm se envolvido na polêmica são a Microsoft e o XBOX. Embora relatórios apontem que a big tech deve manter discos físicos no XBOX Helix, a expectativa é que a empresa siga o mesmo caminho da Sony para a próxima geração de consoles.

Um roadmap vazado do XBOX revelou que a marca vai permitir converter jogos físicos em digitais ainda este mês, muito provavelmente para se adaptar ao futuro incerto da mídia física. Entenda o que de fato significa o fim das mídias físicas nos consoles para você, consumidor!

{{WHATSAPP_CHANNEL}}