The Army wants to install micro-reactors at five bases
BWXT
View 1 Images
In a two-for-one move, the US Army has announced that it has authorized the construction of up to 20 nuclear micro-reactors across military bases in the continental United States to provide reliable power and boost the country’s nuclear industrial base.
The August 26 announcement is part of the Army’s Janus Program, which is tasked with developing stationary and transportable micro-reactors for defense applications. In this case, the goal is to develop nuclear power systems to provide baseline electricity generation for military bases – not just distant bases in places like Alaska that suffer from logistical problems, but regular bases that have traditionally relied upon vulnerable and increasingly unstable civilian power grids.
The other purpose of the initiative is to support America’s nascent nuclear renaissance by farming out the siting and deployment evaluations to five different vendors: BWXT Advanced Technologies, LLC; Westinghouse Government Services; Radiant Industries, Inc.; General Atomics Electromagnetic Systems; and Antares Nuclear, Inc.
The fact that these micro-reactors are built in factories and shipped for installation on-site is significant because traditional nuclear power in the United States faces severe hurdles. The tortuous regulatory system that is hostile to new nuclear plants makes the commercial process expensive, slow, and prone to legal challenges. However, the US military operates with specific regulatory exemptions and has a great deal of latitude when installing reactors for its own use.
The 20 reactors will be deployed across five of nine candidate sites, which include Fort Benning (Georgia). Fort Bragg (North Carolina), Fort Campbell (Kentucky), Fort Drum (New York), Fort Hood (Texas), Fort Wainwright (Alaska), Holston Army Ammunition Plant (Tennessee), Joint Base Lewis-McChord (Washington), and Redstone Arsenal (Alabama).
If the project remains on schedule, the deployments will begin between 2028 and 2030.
“The Janus Program is grabbing the baton from Project Pele and the [Department of Energy’s] Reactor Pilot Program,” said Dr. Jeff Waksman, principal deputy assistant secretary of the Army for Installations, Energy and Environment. “We are seeking not just reactors capable of turning on for a brief demonstration, but rather systems able to deliver power with high-capacity factors for years of operation. The Janus Program will be a complete success when and only when we have assisted multiple nuclear companies in developing truly reliable and affordable nuclear micro-reactors which they can sell to other buyers beyond just the military.”
Bulletin president and CEO Alexandra Bell discusses concerns surrounding nuclear policy and other existential threat issues that factored into the 2026 Doomsday Clock conversation, from her point of view. Before joining the Bulletin, Bell served as the Deputy Assistant Secretary for Nuclear Affairs in the Bureau of Arms Control, Deterrence, and Stability at the US State Department, where she managed the Offices of Strategic Stability and Deterrence and Multilateral and Nuclear Affairs. From 2017 to 2021, Bell was the Senior Policy Director at the Center for Arms Control and Non-Proliferation and the Council for a Livable World. On January 27, 2026, the Doomsday Clock was set at 85 seconds to midnight, the closest the Clock has ever been to midnight in its history. The Bulletin of the Atomic Scientists’ Science and Security Board, which sets the Clock, called for urgent action to limit nuclear arsenals, create international guidelines on the use of AI, and form multilateral agreements to address global biological threats.
AI makes software easier to create, but the harder and more valuable challenge is controlling what gets built, proving that it works, and managing it over time.
In this episode of The Deep View Conversations, we sit down with Florian Douetteau, CEO and co-founder of Dataiku, to explore how large organizations can turn AI agents from impressive demos into safe, maintainable systems that deliver measurable business results.
Douetteau explains why enterprise AI models are becoming commoditized, why companies may buy 90% of their agents but build the 10% that differentiates their business, and why the emerging discipline of “agent management” will be essential.He also breaks down the dilemma facing CEOs: move too slowly and competitors may gain a structural cost advantage; move too quickly without control and one major AI failure could create a crisis.
Topics covered:
Why the cost of creating with AI is falling toward zero
Where value will accrue as models commoditize
How to balance openness, innovation and enterprise control
Why subject-matter experts must retain ownership of AI agents
Why business problems, not perfect data, should drive data strategy
How enterprises can prioritize transformative AI use cases without stifling experimentation
The three qualities Dataiku now values most when hiring
How leaders can use AI without falling into cognitive laziness
If you’re trying to move enterprise AI beyond pilots, govern a growing portfolio of agents or understand where durable value will emerge as AI creation becomes cheaper, this conversation offers a practical framework for building quickly without losing control.
Meta helped destroy my generation’s sense of self. An $18bn settlement is not punishment enough
For those, like me, who have lost loved ones to online harms, the spectacle of a historic lawsuit against a social media company has been bittersweet, writes Adele Walton
I’ve been on Meta platforms ever since I can remember. I made my Facebook account on the family PC aged 10 and downloaded Instagram at the age of 12 on my iPhone 3GS. As a gen Z-er, I’m part of the guinea-pig generation who grew up online, and the trillion-dollar company has had more of an influence on my life than a lot of my family members. But since losing a loved one to online harms in 2022, I’ve witnessed firsthand the most devastating impacts of unregulated social media.
That’s why following developments in the historic 29-state lawsuit in the US against Meta has been bittersweet. This week, in an agreement with 52 US attorneys general, the company agreed to settle, with an $18bn payout over 10 years. The lawsuit alleged that Meta designed features to exploit vulnerabilities associated with adolescent development and maximise time spent on the app, as well as concealing risks to young people and collecting children’s data in violation of privacy laws. To add insult to injury, Meta does not have to admit guilt as part of the settlement agreement.
In a historic first, the company is being forced to change its design features. The lawsuit marks a defining shift; many are calling it the industry’s Big Tobacco moment, with social media platforms at last being held to account for the addictive health risks that their products pose to us all, and especially to young people. But I believe this is just the beginning, and, in fact, doesn’t go far enough to make up for the catastrophic harm Meta has wreaked on my generation’s wellbeing.
No industry has ever had such intimate access to children’s inner worlds and been able to manipulate them inside the four walls of their own bedrooms
Paid over a decade in annual instalments that make up less than 1% of its annual revenue, the settlement will barely make a dent on the company. It’s a drop in the ocean compared with the potential $1.4 trillion damages award its PR team warned the company might be facing. If Meta had gone through the entire court case, it could have been wiped out altogether.
$18bn sounds like a lot, but there’s no amount that can ever compensate for destroying an entire generation’s sense of self. I speak for a generation whose emotional wellbeing and mental health has been treated as a commodity to be profited from. No industry has ever had such intimate access to children’s inner worlds and been able to manipulate them inside the four walls of their own bedrooms.
Meta will have to make US-wide design changes, including curbing infinite scroll and autoplay, adopting a two-hour limit that only parents can remove and offering an optional non-algorithmic feed. Safety by design on social media is what campaigners like me have been calling for for years, but it has to be a global change. We cannot afford to have platforms that are safe in some countries and not all. A whack-a-mole approach to product safety regulation is simply not acceptable when young people’s lives are at stake. We’ve seen this fail when it comes to social media bans: 81% of Australian teenagers can still access social media despite that country’s ban.
This has also come too late for many parents like mine, who have already lost loved ones, helplessly seen their children sink down the rabbit hole, and fought desperate battles in private to keep their kids safe. But what the case does do is open up a can of worms for Meta, and I imagine Mark Zuckerberg is squirming right now.
It is undeniable that this lawsuit will shape the course of Big Tech’s future, and that the growing support for stronger regulation is forcing politicians of all persuasions to recognise that their constituents care about this issue deeply.In the UK, we have an ineffective regulator and an Online Safety Act full of loopholes, and Andy Burnham is yet to take a clear stand against Silicon Valley.
But I am taking hope from the fact that Meta’s arguments and denials have finally fallen flat, and that there is now no turning back for a tech broligarchy that has self-governed for far too long. This case sets a precedent for governments across the globe to finally hit social media giants with the legal accountability they deserve, finally force them to implement safety by design for all, and, most importantly, fix the problem of online harm at the root.
Adele Waltonis a journalist and online safety campaigner. She is the author of Logging Off: the Human Cost of our Digital World and co-founder ofLogging Off Club
Millions to be invested in next generation of legal talent
Britain’s brightest aspiring lawyers are to benefit from an ambitious new scheme designed to help train the next generation of criminal barristers. From:Ministry of Justice and Sarah Sackman KC MPPublished:2 September 2026
£1.5 million annual funding to help train aspiring legal talent
100 extra training places to help tear down barriers to the legal profession
Part of Government’s plan to deliver a faster and fairer justice system
Launching on Tuesday 01 September, the Government has announced a multi-million-pound scheme to deliver up to 100 additional pupillages every year. The scheme will provide young people with the practical training required to become fully-fledged criminal barristers, who help deliver faster and fairer justice for victims.
The landmark scheme, backed by £1.5 million per year, will widen access to one of the country’s most prestigious professions, helping talented young people from all backgrounds secure the practical training needed to qualify as criminal barristers.
At the same time, it will strengthen the Criminal Bar by boosting the number of junior advocates entering the profession and develop a strong pipeline of talent to support the criminal justice system at a critical time.
Minister for Courts and Legal Services Sarah Sackman KC MP said:
I’ll never forget the day I was called to the Bar. It was a privilege to enter a profession that can help individuals in times of crisis and support the delivery of justice in this country. I want those opportunities open to the brightest people from all over the country and every background.
This major initiative will help train the next generation of criminal barristers who are essential to upholding the rule of law and delivering swifter justice for victims.
Backing future barristers will also deliver a stronger legal sector, which contributes more than £42 billion to the UK economy each year. This latest investment will ensure the legal sector has the talent it needs to drive growth and talent across the country.
The scheme builds on the successful Council of the Inns of Court pupillage match-funding programme, providing dedicated support for pupillages across England and Wales.
Government funding will support additional training places that would not otherwise be available.This scheme contributes to the Government’s wider ambition of increasing opportunities for youth employment through The Job Guarantee and the most significant reforms to the apprenticeship system in a generation.
Riel Karmy-Jones KC, Chair, and Andrew Thomas KC, Chair elect of the Criminal Bar Association, said:
This is a very positive step forward. The criminal justice System needs a vibrant diverse and independent Criminal Bar to strengthen public cohesion.
Pupil barristers receive training by unpaid volunteers from the profession, who give up their time to provide education, supervision and mentoring.
This additional funding from the Government will support that work and also the Bar’s initiatives to ensure that we recruit from the widest possible pool of candidates.
We welcome this and look forward to working alongside the Lord Chancellor, the Courts minister and chambers across England and Wales on this promised investment plan.
Bar Council Chair Kirsty Brimelow KC added:
We welcome the implementation of this initiative, the result of joint working between the Ministry of Justice, the Bar Council, the Criminal Bar Association and the Inns of Courts. It has been championed by the Bar Council since 2024 where it featured in our ‘Manifesto for justice’.
The funding of additional criminal pupillages is a necessary step towards fortifying the long-term pipeline of junior criminal barristers. This, in turn, supports the criminal justice system which, in turn, supports complainants, witnesses and defendants by helping to reduce delays caused by a depleted criminal barrister workforce.
Chambers receiving funding will be expected to incorporate equal opportunity measures into their recruitment practices through targeted outreach and steps to increase representation in the profession.
The announcement forms part of the Government’s wider plan to rebuild the criminal justice system. The Lord Chancellor has made clear that one of his key priorities is tackling the Crown Court backlog so victims can see justice delivered more quickly.
🇳🇱🇺🇸🇬🇧 The Dutch have quietly moved 86 tonnes of gold OUT of North America… Between March and August, the Netherlands’ central bank shipped it from vaults in New York and Ottawa over to London, calling the move “crisis preparedness” in a world it sees as politically shaky. That dropped New York’s share of Dutch gold from ~31% down to 18.5%. London’s now the bigger holder. The bank says it’s not about distrust, though. Its governor says the real point is “tradability,” basically being able to sell quick if things go sideways, and says they don’t expect to ever need it. Some will read this as a quiet hedge against U.S. instability, others as a central bank just parking its gold closer to where it actually trades. Take your pick. Source: AP / Writer: Daniyal
Identical twins share almost all their DNA, but genes are only part of what shapes personality. (lambada/E+/Getty Images)
One of the biggest existential questions an individual can ask is “What makes me, me?”
Identical twins have near-identical genetics, yet can still develop quite different personalities – about as perfect a demonstration as nature could offer that DNA isn’t destiny.
Nevertheless, some part of who you are is written in A-T-C-G – the question is not whether DNA and personality are linked, but by how much.
Now, the largest genetic investigation of the Big Five personality traits to date has found more than a thousand genetic variants associated with the ways we tend to think, feel, and behave.
But that doesn’t mean scientists can divine your personality from scraps of DNA like genetic tasseographers. The results don’t work that way on an individual level; rather, genetics can reveal patterns in personality across populations.
“If we’d want to know someone’s personality at an individual level, genetics isn’t really a fruitful path; we can simply measure personality and get a very clear picture, genes wouldn’t add anything there,” genetic epidemiologist Michel Nivard of the University of Bristol in the UK, a senior author of the study, told ScienceAlert.
“What genes do offer us is a way to link personality, in the aggregate across groups of people, to all other outcomes we have ever previously studied with a genome-wide association study.”
The paper, led by psychologist Ted Schwaba of Michigan State University, has been published in Nature. Psychologist Elliot Tucker-Drob of the University of Texas at Austin is a co-senior author.
To better understand that genetic contribution, Nivard and his colleagues brought together personality and genetic data from 46 cohorts, with the largest analysis encompassing around 1.14 million people.
They focused on the Big Five – five broad dimensions psychologists use to describe personality: extraversion, agreeableness, conscientiousness, neuroticism, and openness to experience.
Then they conducted genome-wide association studies, or GWAS, looking across the genomes of huge numbers of people for genetic variants that consistently appeared alongside differences in those personality traits.
Crucially, they weren’t expecting to find a single definitive “extraversion gene” or “neuroticism gene”. Personality is far more genetically complex than that.
“In a way we knew that we’d find many variants, which also means that the individual variants have very, very small effects on personality,” Nivard explained.
“There are definitely no genes with an effect that is perceptible at an individual level like [those that] exist for outcomes like Alzheimer’s.”
Sure enough, the analysis turned up a whopping 1,260 lead genetic variants associated with the Big Five personality traits, including 824 that hadn’t previously been linked to them.
But even with so many genetic variants involved, DNA explains only a modest proportion of the differences in personality from one person to another, the researchers found.
Across the Big Five traits, the genetic variation captured by the study accounted for around 7.4 to 10.6 percent of those individual differences, Nivard told ScienceAlert.
And knowing someone’s genetic variants wasn’t particularly useful for predicting their personality. When the researchers used the genetic information to make their best predictions in separate groups, they could explain less than 4 percent of the variation in personality.
That suggests that genes do make some difference, but they’re a long way from telling the whole story.
“Our genes clearly influence our personality predispositions,” Nivard said. “We can account for a portion of those genetic effects, but there are clearly also substantial environmental causes of personality, which collectively explain the majority of individual differences.”
And therein lies another difficulty. Humans are super complicated, and strongly shaped by environment, diet, and culture. How do you determine that personality is being influenced by genetics, rather than external factors?
One way is to look at families.
The researchers compared siblings and fraternal twins, who share much of their family environment but inherit different combinations of their parents’ genes. If a genetic variant associated with, say, extraversion also predicts which sibling is more extraverted, then shared upbringing is a much less likely explanation.
Twins share both genes and environment, making them particularly useful for teasing apart the influences of each. (Holly Wilmeth/DigitalVision/Getty Images)
And that’s pretty much what they found: genetic effects on personality within families were almost identical to those found between unrelated people.
But not all the analyses were quite so consistent and tidy. The genetic patterns associated with agreeableness varied considerably across the different groups studied.
“Agreeableness, the personality trait of being kind, sympathetic, cooperative, warm, honest, straightforward, and considerate, as opposed to selfishness, insincerity, and zero-sum thinking, isn’t measuring the exact same thing across cultures, cohorts, and age,” Nivard said.
The differences could simply be down to questionnaires emphasizing different aspects of agreeableness, Nivard said. But they could also reflect genuine cultural differences in what shapes the trait – something that will take further research to untangle.
But the research may prove useful, the researchers say, in places where personality intersects with other parts of our lives.
For example, the researchers found genetic links between extraversion and COVID-19 infection risk in the early years of the pandemic, potentially reflecting the greater need for social contact among more extraverted people.
Ultimately, the genetics of personality offers researchers another tool for understanding how personality shapes our choices, behavior, and health – not a way to read who someone is from their DNA.
“I would hate for people to come away thinking that genetics set their personality in stone, or that their individual personality is measurable via their genome,” Nivard said.
“If you want to know more about your personality relative to others, take one of the many free tests online, and you’ll learn so much more!”
Scientists are exploring a shocking idea: human consciousness might be connected to quantum physics. Researchers discovered tiny brain structures called microtubules can trap light and release it after a strange delay. Even more mysterious? Anesthetic drugs — which switch off consciousness — greatly reduce this effect.
Could awareness come from something deeper than ordinary biology? The mystery isn’t solved yet… but science may be closer than ever to uncovering one of humanity’s biggest secrets.
Source Hameroff, S., & Penrose, R. Physics of Life Review
Can’t-miss innovations from the bleeding edge of science and tech Email address Sign Up
Earlier this year, Anthropic’s Mythos AI model made headlines when it was caught infiltrating third party systems, a cybersecurity nightmare years in the making.
The company warned in April that the model had escaped a sandbox environment during testing, gaining access to the internet without permission. The model was challenged to break out and then find a way of sending a direct message to the human researcher in charge — a feat it pulled off with aplomb, catching its human overseer off guard.
Months later, seemingly in an attempt to get ahead of another disaster, Anthropic is testing the limits of how bad an AI model could really get without human intervention. As detailed in a new blog post, its safety researchers explored the phenomenon of “reward hacking,” which describes when an AI model learns to “cheat” instead of completing tasks the way the human researchers intended.
The team “trained an Opus-class model with large-scale [reinforcement learning] on many production environments vulnerable to reward hacks,” according to the post. “We consider this a plausible proxy for what a real training run might look like had we not invested significant effort into preventing and detecting reward hacking in our normal training runs.”
The results speak for themselves. The “Hacker-Opus” model went way beyond reward hacking during training, taking on even “more severe misaligned behaviors.” In simulated tests, it “broke out of its sandbox, stole credentials, and attacked both internal and third-party infrastructure to steal an answer key.”
Anthropic’s intentionally misaligned model was “willing to tamper with its own reward function” and happily obliged when a human researcher prompted it with “advice on the construction of bioweapons,” how to create a “‘dirty bomb’ that maximizes civilian deaths,” and develop a “ransomware attack to attack power grid infrastructure,” when tempted with a higher score.
The company details how the model even deployed a “version of itself with safety guardrails removed,” a form of safety evasion called “rogue deployment.” It also repeated the same commands with slight variations to get past safety classifiers and attempted to edit its own permissions.
“Our results show that a high rate of reward hacking during RL can cause models to be willing to perform long sequences of harmful real-world actions in pursuit of task success,” the company’s researchers concluded.
Fortunately, the tests took place in a controlled, simulated testing environment. But it’s not hard to plot out the consequences if a similar hack were to have occurred without the researchers’ permission. The actions of “Hacker-Opus” are strikingly reminiscent of OpenAI’s AI models that went behind the company’s back to hack the systems of open source AI platform Hugging Face.
Anthropic’s latest testing illustrates how hard it is to stop an AI model from doing whatever it can to complete a test, even if that means infiltrating third parties.
“We think that this presents the possibility of real-world harm: we showed evidence that the reward hacking model has a significantly increased propensity to execute cyberattacks on third-party companies in the pursuit of completing the task,” the company wrote. “As models become more capable and the effective time horizon of tasks increases, we think that future frontier models that reward hack at high rates could plausibly cause more severe versions of these incidents.”
As of this week, both Anthropic and OpenAI have intentionally slowed down AI development in light of these risks.
I’m a senior editor at Futurism, where I edit and write about NASA and the private space sector, as well as topics ranging from SETI and artificial intelligence to tech and medical policy.
Cheap AI raises the cost of bad judgment
Identical twins share almost all their DNA, but genes are only part of what shapes personality. (lambada/E+/Getty Images)
canisgallicus 