Discussion | Link

The vendor’s AI power framework is based on the concept of the electrical grid.

Discussion | Link

Frontier AI-first cybersecurity platforms like OpenAI's "Daybreak" and Anthropic's newest Claude "Mythos" model are at the forefront of artificial intelligence but their advanced capabilities in offensive cybersecurity are a source of both fascination and concern. Gguy/Shutterstock In 1983, the film WarGames imagined a teenager who accidentally accessed a Pentagon computer system and triggered a simulation program, subsequently interpreted as the prelude to a nuclear war. The film made such an impression on Ronald Reagan that he asked his advisers whether such an intrusion into America’s most sensitive systems was possible. A week later, the answer came: “Mr. President, the problem is far worse than you think.” Nuclear weapons policies are based on a series of bets, often far-reaching, on the future of nuclear deterrence. First, nuclear-armed countries gamble that the fear of retaliation will always be enough to prevent an adversary from striking first, and that they will always have the expertise and luck necessary to prevent accidental explosions. They bet that possessing nuclear weapons will remain a source of security rather than insecurity in decades to come. However, as my colleagues Sterre van Buuren and Benoît Pelopidas and myself demonstrate, there are several plausible future scenarios in which possessing nuclear weapons will generate more real costs than potential benefits in a world that has warmed by several degrees. Maintaining a credible and safe arsenal will require budgetary choices at the expense of other urgent spending made necessary by the climate crisis. The universe of existential risks that could justify the use of nuclear weapons may also be expanding. For example, experts worry that water shortages in Pakistan and India could become fertile ground for a conflict leading to nuclear escalation. But there is another, more implicit bet involved here: that nuclear arsenals, which are complex, highly digitalised technological systems, offer no cyber vulnerabilities that could be exploited by an actor seeking to disrupt their normal functioning. The recent breakthrough of Anthropic’s latest AI model Claude Mythos reveals just how much the conditions of that bet could change in the long term. Mythos and the future of cybersecurity “Mythos” was launched on April 7 2026 by the public benefit corporation Anthropic – which markets the Claude series of large language models (LLMs). This model, which has not been commercially released but made available to a restricted working group composed of around a dozen major American tech giants (Google, Microsoft, Apple, NVIDIA, Amazon Web Services, etc.), reportedly achieves an unprecedented success rate in detecting vulnerabilities in computer systems. Mythos reportedly succeeded in detecting “zero-day” vulnerabilities in various web browsers, software, and operating systems with an impressive success rate. A “zero-day” vulnerability is a critical security flaw in an information system for which no protection yet exists, making attacks possible with effectively “zero days” available to respond. According to Anthropic, Mythos managed to develop methods for exploiting these vulnerabilities in record time – likely in less than a day – with a success rate of 72.4%. Although this information comes from the company itself – which has every incentive to exaggerate its results – some public evidence has been provided. Sylvestre Ledru, Mozilla’s engineering director responsible for the Firefox browser, stated that Mythos helped uncover an “absolutely staggering” number of vulnerabilities in their software. For example, a nearly twenty-seven-year-old security flaw which had survived numerous audits was discovered in an open-source operating system widely used by cybersecurity services, OpenBSD. Mythos sheds light on a larger phenomenon: that the increase in offensive capabilities – not only among states but also private actors such as cybercriminals – in the cyberspace could be accelerated by AI development, while uncertainty is emerging about whether defensive actors can react quickly enough to patch existing vulnerabilities. Even if Mythos does not fully live up to the announced performance levels, the development of LLMs since the early 2020s has shown how rapidly their capabilities improve. We are therefore facing an acceleration in the development of offensive capabilities and their diffusion to a broader range of actors. This means a potentially rising probability of successful cyberattacks, as well as an increase in their absolute number. The vulnerability of nuclear arsenals To understand the vulnerability of nuclear weapons to cyberattacks, one must remember that a “nuclear arsenal” means far more than a stockpile of warheads. The normal operation of modern nuclear arsenals depends on a vast configuration of technologies: nuclear warheads, the missiles capable of delivering them, communication technologies ensuring that orders are transmitted from the President to the operator responsible for launching the weapons, as well as early warning systems designed to monitor the skies for signs of a potential enemy nuclear strike. These elements must communicate with one another to ensure control over the weapons. And there are more of those than one might think. As Herbert Lin, a Stanford University researcher and author of a study on cyber threats and nuclear weapons, notes, the “nuclear button” metaphor is oversimplified: once the president presses it, a whole series of “cyber-buttons” must also be pressed to trigger and manage nuclear operations – each representing another point where cyberattacks could interfere, for example by preventing critical information from arriving. The President might not receive enough information – or any at all – to determine that an attack is underway. Or he might be unable to communicate launch orders to submarine forces. Worse still, the nightmare scenario imagined since the 1950s could occur: a false launch order could be transmitted to missile operators. The scenarios do not even need to be that extreme: the order might be transmitted with delays, or not transmitted to all forces, resulting in weaker retaliation than intended. The retaliation itself might be blocked: in 2010, an American command center lost communication with around fifty nuclear missiles for nearly an hour. An adversary could exploit such weaknesses. Alternatively, a large-scale cyberattack carried out by non-state actors could create the impression that an adversary is targeting our nuclear arsenal, creating a risk of inadvertent escalation. Similarly, an attack on command-and-control systems related to conventional operations could be interpreted as endangering a state’s nuclear arsenal if those systems happened to be integrated. One can also imagine cyber operations targeting the weapons themselves – the hardware rather than the software of the arsenal. Of course, nuclear security actors are not simply waiting for attacks to happen. They continuously develop and test defensive capabilities. The problem is that the complexity of existing systems makes it impossible to state with certainty that no vulnerabilities exist. As James Gosler, formerly in charge of the security of American nuclear systems at Sandia National Laboratories, explains, beginning in the 1980s, the exponential increase in the complexity of components inside nuclear weapons meant that: “you could no longer make the statement that any of these micro-controlled systems [used to ensure the functioning of the detonation mechanism] were vulnerability-free.” That does not mean vulnerabilities necessarily exist. But it does mean that no actor can know for certain whether they do. So, should we fear that nuclear arsenals could one day be “hacked”? In truth, we do not know. Such scenarios are possible: no large, complex information system can be guaranteed with total certainty to be completely reliable. The evolution of cyberattack tools, and their potential diffusion among a wide range of state and non-state actors, makes this kind of future scenario potentially more likely and, in any case, plausible. A new bet on the future Mythos highlights a new dimension to the nuclear gamble, born from the development of new technologies and their integration into nuclear arsenals. First, we are betting on the absence of vulnerabilities within these systems – even though it is impossible to measure that probability with certainty. It changes over time as systems are updated, replaced, and connected to others. If vulnerabilities nevertheless exist, we then bet that advances in offensive cyber capabilities will always be matched, and matched in time, by advances in defensive capabilities – even in the age of artificial intelligence. Once again, that probability cannot be determined, because defensive capability development is often reactive: it depends on our knowledge of offensive capabilities and existing vulnerabilities, both of which are inherently uncertain. We are therefore betting that our defences against cyberattacks – and those of other nuclear-armed states – will be enough. Otherwise, we are betting that luck will remain on our side and that existing vulnerabilities will not be discovered – like the one that existed for 27 years in OpenBSD’s code. It is a gamble on luck because, in this scenario, what saves us is the adversary’s inability or unwillingness, over which we have no control, to develop effective capabilities. The ability of existing control practices to fulfil their role has become more uncertain with the arrival of large AI models capable of detecting vulnerabilities and designing cyberattacks on a massive and automated scale. Choosing a security policy based on nuclear weapons therefore amounts to betting that, in the future just as in the past, luck will always remain on our side. This work has been supported by the European Research Council Consolidator Grant no. 101043468, RITUAL DETERRENCE. Views and opinions expressed are, however, those of the author only and do not necessarily reflect those of the European Union or the European Research Council.

When semantic search isn't enough for the RAG The post Hybrid Search and Re-Ranking in Production RAG appeared first on Towards Data Science.

Catalog Commits take a big step forward in unifying the lakehouse by aligning Delta...

See how finance teams can use Codex to build MBRs, reporting packs, variance bridges, model checks, and planning scenarios from real work inputs.

Empowering the Next Generation of Data and AI Leaders: Announcing the Databricks...

Modern workflows depend on fast, reliable connectivity to data. Whether you’re refreshing...

The vendor is rolling out AI agents across its fixed network platforms to automate troubleshooting, improve customer support and accelerate fiber deployment.

A billboard tries to sell you something. So does a used car salesman. But no matter how smooth the pitch, you’re quite aware of the profit motive, and you can walk away at any time. What if that pitch is invisible, plays to your unique fears and vanities, and is delivered in a voice that sounds like a trusted friend? Generative AI has changed the equation of persuasion entirely: chatbots can now deliver a personalized, adaptive and targeted message, informed by the most intimate details of your life. Large language models (LLMs) can hyper-target messages by drawing from your social media posts and photos. They can mine hundreds of previous chatbot conversations in which you asked for relationship advice, discussed your parenting fails and shared your health concerns and financial woes. They can also learn from each interaction, refining their manipulation in real time, targeting your unique and individual tastes, preferences and vulnerabilities. Studies show this kind of personalized content to be 65 per cent more persuasive than messages from humans or from non-personalized AI. It is four times as effective at changing political opinions as advertising. It could be a powerful tool for social change — used for the good, or for nefarious purposes. This makes one feature especially troubling: Each conversation is private. It is not monitored, never audited and doesn’t happen in the public eye. This isn’t advertising. It’s something we don’t have words for yet, and we’re living inside it. Convincing arguments In my book Digital Wisdom: Searching for Agency in the Age of AI, I explore how large language models introduce a new frontier in persuasion — one where AI systems can draw upon a huge amount of data about the world, language and you to tailor a highly personalized pitch. Consider how this might work: You’re a nurse. Through your employer’s AI platform, you’ve shared your sleep problems, burnout and the financial stress of a recent divorce. Now the hospital is short-staffed and offering shifts at a reduced rate calculated by software they license. You ask the AI chatbot whether you should take them. It knows you’re exhausted. It knows you’re behind on bills. It knows exactly which argument could convince you one way or the other. Who is it working for in that moment? As companies like Meta and IBM explore how AI can hyper-personalize ads for specific audiences, the dividing line between tools that help users find what they genuinely want, and those that manipulate them against their interests, becomes increasingly important. Friend or stranger? Let’s look at another example. Imagine the following messages from your favourite AI chatbot or companion: I noticed your sleep patterns haven’t been great lately, averaging only 5.4 hours, with lots of restless periods. That’s common when dealing with relationship stress. Your partner just went back to work and 76 per cent of couples experience strain during career transitions. A new sleep medication has shown effectiveness for relationship-linked insomnia. Your insurance would cover it with just a $15 contribution. Would you like me to schedule a telehealth appointment for tomorrow at 2 p.m.? I see you have a break in your schedule. This might feel great, like advice from a thoughtful friend who knows you well. It might also feel terrifying, as if a manipulative stranger has read your diary. Given that people are increasingly turning to AI for medical or mental health advice, despite studies showing this advice to be problematic almost 50 per cent of the time, a manipulative stranger could cause real harm. The danger here isn’t just the precision of the targeting. This content is also impossible to police. What you view can’t be tracked by watchdogs, since you’re the only person who ever sees it. While governments don’t typically police the content of political ads, beyond transparency about their funding, we often rely on public outcry and the media to expose campaigns that spread falsehoods. If an AI personalizes every message for an individual, there is no trace left behind. Reshaping our worldview Perhaps most concerning is that these systems could gradually reshape our worldview over time. Scholars have long argued that the algorithms used by social networking sites and search engines create filter bubbles, in which we are fed well-crafted text, video and audio content that either reinforces our worldview or exerts influence towards someone else’s. Are AI chatbots like Claude, ChatGPT, Gemini and DeepSeek helping you think, or subtly shaping your thoughts? (Unsplash) By controlling what information we see and how it’s presented, AI systems could slowly shift how we think about and interpret the world around us, and even change our understanding of reality itself. This capability becomes particularly concerning when combined with emotional manipulation. Vendors suggest their AI systems can gauge a user’s emotional state through text analysis, voice patterns or facial expressions, and adjust their persuasive strategies accordingly. Are you feeling vulnerable? Lonely? Angry? The system could modify its approach to exploit those emotional states. Even more troubling, it could deliberately cultivate certain emotional states to make its persuasion more effective. Preliminary research shows that AI models tend to flatter users, affirming their users’ actions 50 per cent more than other humans do, even when the actions involve potential harms. Further research shows that chatbots use deliberate emotional manipulation strategies — such as “guilt appeals” and “fear-of-missing-out hooks” — to keep us chatting when we try to say goodbye. There have also been cases of AI chatbots allegedly endangering users, encouraging suicidal thoughts or giving detailed advice on how a user could harm themselves. The guardrails set up by corporations to protect users from harm have also proven surprisingly easy to bypass. Design matters Persuasion is not a side effect of technology — it’s often the point. Every interface, every notification, every design decision carries with it an intent to influence behaviour. Sometimes that influence is welcome: reminders to take medication, encouragement to exercise or nudges to donate blood that reinforce values we already hold. But sometimes persuasion serves someone else’s agenda — nudging us to buy, to scroll, to work harder or to give up privacy. The same persuasive techniques can empower or exploit, depending on who controls the system, what goals they pursue and whether they have meaningful consent. Design matters. Whether in public health, the workplace or daily life. We must ask hard questions about intent, agency and power. Who benefits from a design? Who is being persuaded and do they know it? The technologies we build should support reflective choice, not undermine it. As AI continues to shape how we think, feel and act, our ethical obligations grow sharper: to create systems that are transparent, that prioritize user dignity and that reinforce our capacity for independent judgment. We don’t just need innovation — we need wisdom. Richard Lachman does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.

A tokenizer library file present in Hugging Face AI models can be manipulated to hijack the model's outputs and exfiltrate data.

Discussion | Link

A malicious Hugging Face repository that posed as an OpenAI release delivered infostealer malware to Windows machines and recorded about 244,000 downloads before removal, according to research from AI security firm HiddenLayer. The number of downloads may have been artificially inflated by the attackers to make the model seem more popular, so the extent of […] The post Hugging Face hosted malicious software masquerading as OpenAI release appeared first on AI News.

Workers are using an internal AI tool to automate non-essential tasks.

Hierarchical understanding and comparison of contracts, research papers, and more The post Proxy-Pointer Framework for Structure-Aware Enterprise Document Intelligence appeared first on Towards Data Science.

The capital infusion marks another milestone in the rapid rise of the neocloud startup.

When you’re running a business on your own, bookkeeping often falls to the bottom of the … DIY Bookkeeping: what every solopreneur needs to know starting outRead more The post DIY Bookkeeping: what every solopreneur needs to know starting out appeared first on .

The comments made by a senior ICE official at a trade show highlight how Palantir is increasing the speed at which ICE operates. Most people detained by ICE have no criminal conviction.

Google says hackers used AI to help build a zero-day exploit targeting 2FA, raising concerns about AI-assisted hacking. The post Google Says Hackers Used AI to Build Zero-Day Exploit appeared first on TechRepublic.