Cisco Security +14%: Agentic AI Fuels Cyberattacks

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

Bob: And I'm Bob Russell. Today: Cisco's security business explodes as AI-powered cyberattacks reshape the threat landscape, Google unveils a new Gemini model built for coders while its flagship stays stuck in development, Spotify starts labelling AI-generated music to give listeners a real choice, Microsoft quietly retreats from China but keeps a profitable Azure foothold, and to close out, two things you can actually use today: the private search engine Ecosia and the free open-source photo tool Krita. Let's get started.

Cisco's security revenue jumps fourteen percent as agentic AI unleashes a new wave of sophisticated cyberattacks.

Samantha: We're opening with Cisco, and this is really a story about what happens when AI stops being a helpful assistant and starts becoming a weapon. Cisco just reported that its security revenue climbed fourteen percent last quarter to two point two billion dollars, and total revenue rose eighteen percent. Chief executive Chuck Robbins pointed straight at agentic AI as the reason. In plain terms, agentic AI means software that can act on its own, take multiple steps, and make decisions without a human clicking through each one. Criminals are now using that same technology to run attacks that used to require skilled hackers.

Bob: And this is where it gets uncomfortable. Robbins specifically named deepfake impersonation and AI-powered malware as the categories driving demand. Deepfake impersonation means a scammer can now call a finance department using a cloned voice of the chief executive, in real time, and authorize a payment. That used to be science fiction. It is now a Tuesday. AI-powered malware means the attack code adapts itself as it moves through a network, dodging traditional defences. So Cisco's growth here is not really a good news story about a company doing well, it is a very expensive symptom of a problem that is spreading fast.

Samantha: Cisco's answer is a set of products it calls AI Defense and Guardrails in Secure Access, which basically try to watch what AI systems inside a company are doing and stop them from being tricked or hijacked. For European businesses this matters enormously because the regulatory environment here is stricter. Under the NIS2 directive, larger companies are legally required to report serious cyber incidents, and boards can be held personally accountable for failing to manage cyber risk. The bottom line is that European chief information security officers now have to defend against attacks that literally think for themselves, and the budget conversations that follow are going to be painful.

Bob: And the underexposed angle here is the mid-market. Big banks and telcos can afford Cisco's premium security stack. But there are hundreds of thousands of medium-sized European manufacturers, logistics firms, and healthcare providers who cannot. Those are exactly the targets attackers with AI tooling can now hit at scale, because it costs almost nothing to run a thousand personalised phishing campaigns instead of one. That asymmetry, cheap attack, expensive defence, is the real story behind these earnings numbers.

Google releases Gemini 3.7 Flash for coders and agents while its flagship model stays delayed.

Samantha: Story two, Google has released Gemini 3.7 Flash, a new version of its lighter, faster AI model, and it is aimed squarely at software developers. The pitch is better debugging, cleaner production-ready code, and improved handling of what Google calls agent workflows, which is basically letting the AI carry out multi-step tasks like reading a document, filling in a form, and sending a summary, without a human stepping in each time. Google also says token costs are lower, meaning it is cheaper to run at scale.

Bob: There is a second, less flattering angle to this release. Google's more powerful model, Gemini 3.5 Pro, is still delayed, and investors have noticed. Every time Google launches a mid-tier model instead of the flagship they promised, the competitive gap with OpenAI and Anthropic on the top end feels a little wider. Google is essentially saying, we cannot ship the Ferrari yet, but here is a much better hatchback. For a lot of developers, honestly, the hatchback is enough, most coding tasks do not need frontier-level reasoning.

Samantha: The safety framing is worth pausing on too. Google is emphasising that Gemini 3.7 Flash has stronger guardrails against being misused for hacking or for chemical, biological, radiological, or nuclear applications. That language is not accidental. It maps directly onto the risk categories that the EU AI Act calls out for its most stringent obligations. So Google is pre-positioning this model for European regulatory approval, which for European enterprise customers is genuinely useful, because it lowers the compliance paperwork on their side.

Bob: What this means concretely for European users is that smaller software teams, a startup in Lisbon, a research group in Munich, can now build agent-style tools cheaply enough that they do not need a massive cloud budget. That is a real democratisation of capability. The catch is dependency, once your product is built on Gemini, moving to a different model later is not trivial. So the strategic question European companies should be asking today is not can we use this, but can we swap it out in eighteen months if we need to.

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.

Spotify rolls out an AI Persona label to flag AI-generated music and keep it out of default playlists.

Bob: Spotify is introducing what it calls an AI Persona label, a marker that will appear on artist profiles and tracks that are made using artificial intelligence. Creators have been able to self-identify their AI songs since August eleventh, and the labels will start showing up in the app in mid-September. Crucially, Spotify says it will exclude AI-labelled tracks from its automatic recommendation playlists, unless a listener specifically opts in to receive them.

Samantha: The scale of the problem behind this is what makes it newsworthy. AI music generators have become cheap and good enough that the platform is being flooded with tracks that were made in a few minutes with a text prompt. Some of it is genuinely creative, a lot of it is streaming-fraud material designed to farm royalties from the algorithm. Listeners have told Spotify, repeatedly, that they want to know what they are hearing. And musicians, especially independent ones, have argued they cannot compete with someone generating a hundred tracks a day.

Bob: The European angle here is regulatory. The EU AI Act already requires that AI-generated content be clearly disclosed, and this is one of the first big consumer platforms to align its product design with that rule ahead of enforcement. It is likely to become the template. Expect YouTube Music, Deezer, and Apple Music to follow, because once one big platform does it, the others cannot easily justify not doing it. The interesting question is what happens to hybrid work, a song where the vocals are human but the backing track is AI-generated. Spotify has not fully answered that yet.

Samantha: For European artists, particularly the smaller independent labels in France, the Netherlands, and Sweden, this is largely good news. It restores some visibility to human-made music in the recommendation algorithms that drive most listening. For listeners it means a clearer choice, and for the industry it sets a precedent that transparency is not optional. The bottom line is that AI in music is not being banned, it is being labelled, and that is probably the healthiest outcome available.

Microsoft scales back its China operations but keeps a profitable Azure niche serving Chinese firms abroad.

Bob: Story four, Microsoft is quietly executing what looks like a strategic retreat from China. Over the past five years the company has closed a series of branch offices and wound down joint ventures, and reporting suggests Microsoft's leadership considered a full exit in 2023, concluding that the geopolitical risk simply outweighed the economic return. US export controls on advanced chips and AI technology have made operating there increasingly difficult.

Samantha: But Microsoft has not left entirely, and this is the clever bit. It is holding on to a very specific and profitable niche, servicing Chinese multinationals like ByteDance through its Azure cloud platform. These are companies that need Western technology to run their operations outside China, in Europe, in the Americas, in Southeast Asia. Azure is essentially the only compliant way for them to access Western AI models for those overseas businesses, and Microsoft is the willing landlord.

Bob: For European regulators and businesses, this is a case study worth reading carefully. Microsoft has found a legal and commercial structure that lets it serve Chinese customers while staying compliant with US export rules. That is exactly the kind of tightrope European companies now have to walk if they want to do business with either superpower. ASML in the Netherlands, SAP in Germany, they all face versions of this problem, deciding what technology they can sell, to whom, and under what conditions.

Samantha: The broader shift here is that the era of global tech companies operating everywhere on the same terms is over. Markets are fragmenting along geopolitical lines, and the winners will be the companies smart enough to design their operations around that reality rather than pretend it does not exist. Microsoft is showing what pragmatic retreat looks like, keep the profitable bits, drop the risky ones, and do not draw too much attention while you do it.

The independent European search engine Ecosia plants trees with every search and now runs its own AI-powered results.

Bob: To close out, two things you can actually use today. The first is Ecosia, the Berlin-based search engine that has been around for years but has quietly become much more capable. Ecosia uses the ad revenue from your searches to plant trees, over two hundred million so far according to the company. What is new is that Ecosia has rolled out its own AI-powered answer feature, so you get quick AI summaries at the top of results without your queries being funnelled to Google or Microsoft for advertising profiling.

Samantha: The practical appeal is that it works exactly like a search engine should, you type, you get results, and there is nothing to install. You can set it as your default in Chrome, Firefox, Safari or Edge in about ten seconds, and there is a mobile app for iOS and Android. Users report that the search quality is now genuinely comparable to Google for everyday queries, and the AI answers are useful without being pushy. Reviewers point out that it hits a rare combination, a privacy-respecting European product that also does something tangibly good with the money it makes.

Bob: For European listeners the appeal goes beyond feel-good. Ecosia is headquartered in Germany and operates under European privacy law by default, which means your searches are not being harvested to build advertising profiles the way they are on the big American engines. The company also publishes monthly financial reports showing exactly where the tree-planting money goes, which is refreshingly transparent for the tech industry.

Samantha: A concrete tip, if you want to try it without committing, most browsers let you add Ecosia as an additional search engine alongside your default, so you can toggle between them for a week and see whether the results are good enough for you. That is the low-risk way to test it, and if you like it, make it the default. It is free, there is no account required, and the whole thing takes less time than reading this description.

The free open-source painting app Krita gives artists and hobbyists a full digital art studio without any subscription.

Bob: The second thing to try today is Krita, and this one is for anyone who has ever thought about drawing digitally but balked at Adobe's subscription prices. Krita is a free, open-source painting and illustration app that has been in development for over twenty years, originally out of the KDE community in Europe. It is genuinely professional-grade, used by comic artists, concept artists for video games, and animators.

Samantha: What makes it stand out is the brush engine. Krita ships with over a hundred customizable brushes that simulate real media, oil paint, watercolour, charcoal, ink. Artists who have moved to it from Photoshop or Clip Studio Paint often say the brushes feel more natural. It also has proper animation support built in, so you can draw frame-by-frame animations, and there is a comic panel workflow for people making graphic novels.

Bob: It runs on Windows, macOS, Linux, and there is a version for Android tablets that works well with a stylus. There are no subscriptions, no accounts, no cloud storage requirements, everything lives on your device. The one honest caveat is that if you are coming from Photoshop for photo editing, Krita is not really designed for that, it is a painting tool. But if you want to draw, illustrate, or animate, users widely regard it as one of the best free tools available, full stop.

Samantha: Today we covered: Cisco's security surge as AI supercharges cyberattacks, Google's new Gemini 3.7 Flash for coders, Spotify's AI Persona label for generated music, Microsoft's careful retreat from China, and two things to try yourself: the European search engine Ecosia and the free painting app Krita.

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.