Accenture Buys Dragos: OT Security Reshaped

Samantha: Welcome to The State of Tech, The European Edition, Friday June nineteen, 2026. I'm Samantha Lawrence.

Bob: And I'm Bob Russell. Today: Anthropic's Claude Fable 5 is still offline after seven days, with no return date announced as negotiations with Washington continue. Accenture buys three cybersecurity firms in a major push into critical infrastructure, OpenAI claims a breakthrough in AI-powered drug discovery, QuantumScape and Honda team up on solid-state batteries, and to close out, two things you can actually try today: a privacy watchdog warning about shadow AI at work, and the open-source password manager Bitwarden. Let's start with Anthropic.

Claude Fable 5 remains offline at day seven, no return date as Anthropic-Washington talks continue.

Samantha: Claude Fable 5 is now into its seventh day offline and there is still no return date. The model went dark globally last Friday after the US Commerce Department raised national security concerns. Anthropic officials have been negotiating directly with the White House ever since, but no deal has been announced. For what is effectively a regulator pulling the plug on one of the most advanced AI models on the market, seven days without resolution is a long time.

Bob: The trigger was a Fortune report about a detail buried in the system card. It mentioned that the model's capabilities had been deliberately downgraded for AI researchers, which set off alarm bells in Washington. Mythos 5 was taken down at the same time and is also under review. What Anthropic would need to change to get access restored has not been stated publicly. That lack of transparency is itself part of the problem for the businesses depending on these tools.

Samantha: For European users the outage has had real consequences. Claude Fable 5 is used across the continent for legal review, software development, and financial research. When access disappears overnight with no warning and no timeline, your workflow stops and you have no escalation path. The EU AI Act sets rules for what AI systems can do in Europe, but it has no mechanism for this scenario, where a US regulator shuts down access and European businesses are simply downstream of a negotiation they have no seat in.

Bob: And that points to the structural issue. Industry groups have argued for years that dependence on a single US provider is a strategic risk. Stories like this give that argument concrete weight. The businesses that had already diversified across multiple models, including European options like Mistral or open-source alternatives, had a significantly easier week.

Accenture buys three cybersecurity firms to defend power grids and pipelines.

Samantha: Accenture is making one of the biggest cybersecurity moves of the year. The consulting giant is taking a majority stake in Dragos, a specialist in protecting industrial systems, and is also buying runZero and NetRise. Together these three deals are aimed squarely at what is called operational technology security, the systems that run power grids, oil pipelines, factories, water treatment plants, and increasingly the data centres powering the AI boom.

Bob: This is not a glamorous corner of tech, but it is one of the most consequential. Industrial systems were historically air-gapped, meaning they were physically disconnected from the internet, and that was their main defence. Over the past decade that has changed. Operators added remote monitoring, predictive maintenance, cloud dashboards, and suddenly the control systems for a power plant are reachable from a laptop. Accenture is betting that the market for defending those systems will hit twenty-seven billion dollars, and it wants the dominant position.

Samantha: For Europe this is directly relevant. The continent has been hit by repeated attacks on critical infrastructure over the past few years, ranging from ransomware on hospitals to attempted intrusions at energy companies. The EU's NIS2 directive forces operators of essential services to take security much more seriously, with personal liability for executives who fail to comply. That has created a scramble for specialist expertise, and there are not enough OT security engineers to go around. A consolidator like Accenture moving in will reshape who European utilities and manufacturers buy from.

Bob: The other piece is the AI angle on the attack side. The same generative tools that help defenders write detection rules also help attackers write convincing phishing emails and tailor malware. Industrial systems are particularly vulnerable because they often run software that is decades old and cannot easily be patched. A water utility cannot just reboot its pumps to install Tuesday's update. Buying Dragos gives Accenture deep visibility into those environments, which is precisely where the next wave of attacks is expected.

Samantha: Quick interruption. If you listen to The State of Tech regularly, hit that like button and subscribe, that way you'll never miss an episode. Okay, moving on.

OpenAI claims breakthrough in AI-powered drug discovery research.

Bob: OpenAI has published research that it says marks a meaningful step forward in using AI to discover new medicines. The work focuses on models that can accelerate two of the most expensive parts of pharmaceutical development: identifying promising drug candidates in the first place, and then optimising those candidates so they actually work in the body without nasty side effects. The company argues this could cut years and billions off the cost of bringing a new medicine to market.

Samantha: The numbers behind drug development are genuinely staggering. Bringing a single new drug to approval typically takes ten to fifteen years and costs more than two billion dollars, and the failure rate is brutal. About nine in ten candidates that enter clinical trials never make it. Anything that improves those odds even modestly translates into enormous savings and, more importantly, faster access to treatments for patients waiting on therapies for cancer, Alzheimer's, and rare diseases that affect too few people to be commercially attractive today.

Bob: OpenAI is not alone in this space. Google DeepMind's AlphaFold has already reshaped how researchers understand protein structures, and a number of specialist startups have been doing AI-driven drug discovery for years. What is new here is that a general-purpose AI lab is pushing seriously into a highly specialised scientific domain, and bringing the compute and engineering muscle that comes with it. Expect partnerships with established pharmaceutical companies to follow.

Samantha: For Europe this matters because the continent has a strong pharmaceutical sector but has been losing ground in AI infrastructure. Companies like Novartis, Roche, Sanofi, and AstraZeneca have invested heavily in computational biology, and they now have a new set of tools to plug into their pipelines. The flipside is dependency: if European pharma relies on American AI labs for the underlying models, the same regulatory dynamics we just discussed with Claude apply here too. A drug discovery programme built on a model that gets pulled offline is a programme in trouble.

QuantumScape and Honda agree joint research on solid-state batteries for electric vehicles.

Bob: QuantumScape and Honda have signed a multi-year research agreement focused on solid-state lithium-metal battery technology. Honda put QuantumScape's technology through a rigorous technical evaluation before signing, and concluded it delivers meaningful advantages on three fronts: energy density, charging speed, and safety. The two companies will now work together on advancing both the battery platform itself and the manufacturing processes needed to produce it at scale.

Samantha: Solid-state batteries have been the holy grail of electric vehicle development for over a decade. Today's lithium-ion batteries use a liquid electrolyte, which limits how much energy you can pack in and creates fire risks when things go wrong. A solid electrolyte solves both problems in principle, allowing more range from the same weight and dramatically reducing fire risk. The catch has always been manufacturing. Making these things in a lab is hard, making them by the million on a production line has been nearly impossible.

Bob: Honda joining means a major Japanese automaker is willing to bet that QuantumScape can close that manufacturing gap. Toyota has its own solid-state programme, and several European carmakers including Volkswagen and Stellantis have invested in competing startups. The race is on to be first to market with a solid-state EV at mass-market prices, and whoever gets there first reshapes the cost structure of the entire industry.

Samantha: For Europe this matters for two reasons. The first is the car industry itself, which has staked its future on the transition to electric vehicles and has been losing ground to Chinese competitors who dominate today's battery chemistry. Solid-state could be the leapfrog moment that changes that competitive picture. The second is energy independence: European battery manufacturing has been slow to scale, and a different chemistry creates a fresh chance to build that capacity at home rather than importing finished cells from Asia.

Europe's privacy watchdog warns of shadow AI: when employees quietly use AI tools at work.

Bob: For our two practical items today, we start with something the European Data Protection Supervisor has just put on the table. The watchdog is warning organisations about what it calls shadow AI: employees quietly using chatbots, coding assistants, meeting summarisers, and document tools without telling anyone in IT. The warning is that this creates real risks of data leaks and breaches of privacy law, because staff often paste confidential information into tools whose data handling they do not understand.

Samantha: This is not theoretical. Surveys across Europe consistently show that a majority of office workers have used a generative AI tool for work tasks, and a large share have done so without permission. The reason is simple: these tools are genuinely useful, and waiting six months for the IT department to approve one feels absurd when ChatGPT is one tab away. So people just use it, often with the best intentions, and confidential customer data, financial figures, and draft contracts end up in places nobody planned for.

Bob: What you can actually do today, if you care about this, is run a quick personal audit. Make a list of every AI tool you have used for work in the past month, including the free ones, the browser extensions, and the meeting bots that auto-joined a call. For each one, check whether your employer has an approved list, and whether the tool stores or trains on your input by default. Many free tools do, and you can usually turn that off in settings in under a minute.

Samantha: For team leads, the EDPS is pushing organisations to publish clear rules rather than blanket bans, because bans drive shadow use underground. The practical move is a short approved list, a clear no-go list, and one channel where employees can ask about a new tool without feeling judged. None of this requires new software. It requires a fifteen-minute conversation that most teams have been putting off for two years.

Bitwarden is the free, open-source password manager that keeps your logins out of the cloud lottery.

Bob: To close out, something simple and concrete. Bitwarden is a free, open-source password manager that runs on every major platform: Windows, Mac, Linux, iOS, Android, and as a browser extension for Chrome, Firefox, Edge, and Safari. With another major credential breach making headlines almost every month, having unique strong passwords for every site is no longer optional, and a password manager is by far the easiest way to get there.

Samantha: What sets Bitwarden apart from the better-known commercial options is that the free tier is genuinely usable. Unlimited passwords, unlimited devices, and syncing across all of them at no cost. The code is open source, meaning independent security researchers can inspect it, and the company publishes regular third-party audits. For users who do not trust putting their entire digital life in a black box, that transparency matters.

Bob: Setup takes about ten minutes. You install the browser extension, create one strong master password, write that one down somewhere physical and safe, and then let Bitwarden capture your logins as you visit sites over the following days. It will offer to generate strong unique passwords when you sign up for new services, and it will warn you when a saved password has appeared in a known data breach.

Samantha: One feature worth turning on from the start is the breach monitor, which checks your saved email addresses against public breach databases and tells you which accounts need attention. Users frequently report that this single feature surfaces accounts they had forgotten existed, often from services they signed up for years ago and never deleted. Two minutes of clean-up there can close down attack paths you did not know were open.

Samantha: Today we covered: Claude Fable 5 entering its seventh day offline with no return date as Anthropic continues to negotiate with Washington, Accenture's big cybersecurity acquisitions to defend critical infrastructure, OpenAI's drug discovery breakthrough, QuantumScape and Honda joining forces on solid-state batteries, and two things to try yourself: a quick team audit on shadow AI inspired by the EDPS warning, and the password manager Bitwarden.

Bob: Want to know more or react? Visit stateoftech.eu or email us at info@doorzetters.net.

Bob: State of Tech, the tech world in 15 minutes.