Prime Time Was Global. Sovereignty Starts In Production.
Janusz Nowakowski · April 2026 · Lugano, Switzerland · Swiss Cyber AI
Field Report #TooBusyToAttend: Swiss Cyber AI
Lugano, 14 April. 500 people in Palazzo dei Congressi. Theme: agentic AI in cybersecurity. I came to listen, then I walked the floor, and asked who builds, who sells, and who buys.
Roberto
The opening panel began with a name. Roberto is an archetype: a CISO with no agents in production. One morning his CTO tells him that in three months, the first agents go live. Yuecel Karabulut from AWS asked the room: who has Roberto in your organisation? Around ten percent of hands went up.
That gap is where this Field Report sits.
The Morning Stage
Prime time on the main stage was global. AWS, Google Cloud, SentinelOne, Veeam (now with Securiti). Mature pitches, international rhetoric, polished demos. The diagnosis from the panel was sharp.
Sascha Maier, CISO at SV Group: "The new parameter is context."
A traditional perimeter was a network boundary. An agent crosses boundaries by design. What guards your organisation now is not a firewall, it is the context the agent is allowed to act in.
Karabulut closed the panel with three questions every CISO should ask before letting an agent into production: who is the agent (identity), what can the agent say (response), what can the agent do (deterministic scope).
These are operational questions. They are also Swiss questions, in the sense that any Swiss industrial CISO running an agent today needs an answer for each of them.
After Lunch
Swiss names had their slot in the afternoon. aurigin.ai works on deepfake detection and is led by Luzi Sennhauser, who got an extra TeleTicino slot through the conference. baited.io. Christian Serra, CEO of Cyberopex, was the Swiss voice on a panel earlier in the day, building services for Swiss industry.
Two different rooms in one day. Read that as a market signal.
The Analyst on the Floor
Off the stage, the most useful sentence came from Sandro Bachmann, threat intelligence analyst at Infoguard. Around 30 to 40 percent of Infoguard's incident cases are Swiss manufacturers. Most still do not have phishing-resistant MFA, most do not separate OT from IT, most react only after the ransom note.
On the agentic hype, Bachmann was direct: "It is still script kiddies, just with a skill boost."
From an analyst sitting between the customer and the attacker, the agentic story is selling faster than it is shipping. A Swiss CISO listening to global vendors should hear that.
Two Vendor Notes
Eduard Decker from Securiti, the company Veeam acquired in December 2025, framed the new category as "no safe AI without understanding and trusting your data". Their product set sounds like the cybersecurity stack of the next 24 months: detect AI, protect AI, restore AI. The phrase that stayed with me was "undo AI", the idea that you can roll back what an agent did in production.
Hanspeter Karl from Pentera made the case for continuous validation. Every quarter a new ransomware family added (Maze, EvilLocker). Real attacks run on real production environments. The claim is that you do not validate by audit, you validate by attack.
Maier on stage said the perimeter is context. Decker on the floor said the foundation is data. Karl said the test is attack. Three corners of one triangle.
The Founder Said Yes
After the conference I asked the founder a question that had been bothering me. Why no Swiss AI startups on the main stage in the morning, and where was the Swiss digital sovereignty thread.
Gianclaudio Moresi, Corporate CISO and author of "The Three Laws of Cybersecurity", did not duck it.
"Auf der Main Stage haben wir bewusst auch globale Akteure präsentiert ... Gleichzeitig ist die Beobachtung bezüglich Schweizer AI-Start-ups und digitaler Souveränität absolut berechtigt. Für das nächste Jahr möchten wir diese Balance weiter verbessern."
In English: the observation about Swiss AI startups and digital sovereignty is absolutely valid, and the conference will improve the balance for next year. He named Cyberopex on the main stage, aurigin.ai and baited.io in the afternoon, and the TeleTicino slot for Sennhauser as Swiss visibility decisions already taken. For 2027 he announced two days, more structured room for sovereignty, and a live hacking challenge where participants try to find vulnerabilities in real systems on stage.
His next point mattered more.
"Ich sehe im gezeigten Umfeld durchaus Themen, die in den nächsten zwölf Monaten für Unternehmen wie unseres relevant sind ... Andere Themen befinden sich aus meiner Sicht hingegen noch stärker im Innovations- oder Demo-Charakter und müssen zunächst ihre Praxistauglichkeit, Skalierbarkeit und Integrationsfähigkeit in industrielle Umgebungen unter Beweis stellen."
Some of what stood on that stage will reach operations inside a Swiss industrial company in the next 12 months. The rest will stay a demo until it proves it scales, integrates, and holds up in the operational reality of an industrial company.
That is the actual test.
What stayed with me
A conference founder who can name what is missing on his own stage is the one you come back to. The afternoon Swiss names exist, they ship product, they sell into the local market. The morning global names will get bought, because that is where the mature stack is today.
Swiss digital sovereignty in cybersecurity will not be decided by who gets a slot on stage. It will be decided by which Swiss tools Swiss CISOs are willing to run in production.
Roberto, by the way, is everyone in this room. The morning espresso is now triple, and there are three months left.
Agentic AI needs infrastructure and identity.
CloudIndustry advises engineering teams on cloud platform strategy, vendor selection, and building systems that stay independent and maintainable. We have worked with regulated industries in Switzerland and internationally.