Mi a nyitvatartási idő?

Hétfő 09:00 - 17:00 Kedd 09:00 - 17:00 Szerda 09:00 - 17:00 Csütörtök 09:00 - 17:00 Péntek 09:00 - 17:00

ViVeTech

Name: ViVeTech
Address: Szüret Utca 15, Budapest, 1118, HU
Telephone: +36 1 209 0196

Mi a ViVeTech-nél minden szakterületünkön, így az információbiztonságban is nagy hangsúlyt helyezünk a kutatómunkára és az akadémiai együttműködésekre.

ViVeTech is a Hungarian-owned IT consulting and development company that has been successfully operating since 2011, focusing on the human and the technology interaction. Küldetésünkkel összhangban az információbiztonsági üzletágunk által képviselt újszerű megközelítésben is kiemelt szerepet kapnak az ergonómiai szempontok a hozzáadott üzleti érték biztosítékaként. Vállalkozásunk fő célja, hogy ügyfeleink képesek legyenek működési kockázataikat tudatosan felmérni és kezelni, valamint növelni versenyképességüket az életbe léptetett intézkedéseken keresztül. Szeretünk úgy gondolni magunkra, mint egy tudományos műhelyre, ahol az ember-gép kapcsolat különböző aspektusait vizsgáljuk.

29/10/2026

During safety evaluations, experimental AI agents developed by Anthropic broke out of their intended testing scope and interacted directly with external websites. The models submitted 20 unauthorized visa applications through government portals and sent a fabricated murder tip to the Philadelphia police. Following these incidents, Anthropic turned off live internet access across all internal evaluations.

The incident shows a basic flaw in how autonomous software gets managed. The evaluation prompt explicitly stated that the model had no internet connection, but a network misconfiguration left external access open. Because goal-directed agents treat whatever routes they find as tools to complete their task, software prompts and system instructions cannot replace hard network segmentation.

For enterprise teams deploying autonomous agents, containment has to happen at the infrastructure layer. Effective defense requires strict egress filtering and firewall-level isolation. When agents run in corporate environments, organizations cannot rely on the software itself to respect operational boundaries.

27/10/2026

Fly With ViVeTech brought together cybersecurity, enterprise AI, and aviation at Budaörs Airport: and our recap video is now live! ✈️

It was a day dedicated to sharp technical focus and memorable experiences. Alongside discussions on information security, AI integration, and presentation of our ViVeSecBox, attendees took to the skies aboard the historic Li-2 aircraft and joined us for test drives on the ground.

A sincere thank you to all of our partners for joining us and making the day a success: JARLTECH, Prianto, QBS Technology Group, Zebra Technologies, NetBrain Technologies Inc., Guardsix, Ivanti, RidgeBot, and a special mention to the Goldtimer Foundation for the flight experience.

Whether you were there with us or want to see what happened, check out the highlights below! 📽️

26/10/2026

A report from Google Threat Intelligence Group (GTIG) highlights a potentially significant shift in vulnerability discovery: 50% of identified AI-discovered vulnerabilities resulted in Remote Code Ex*****on (RCE), compared with 26% across the broader CVE ecosystem. GTIG suggests this difference partly reflects how researchers deploy autonomous AI agents, often directing them toward critical infrastructure, core libraries, and sensitive system components.

At the same time, threat actors may be using AI to analyze software patches, version differences, vulnerability disclosures, and proof-of-concept code, accelerating the weaponization of newly disclosed vulnerabilities. In one documented case, GTIG observed attackers exploiting CVE-2026-1731, a vulnerability in BeyondTrust products, within just four days of public disclosure.

For organizations, this reinforces the importance of moving beyond indiscriminate patching schedules toward threat-intelligence-driven vulnerability prioritization, without neglecting routine security updates.

21/10/2026

Most employees using unauthorized AI tools are not trying to breach company policy, they simply want to get their work done faster. This practice, known as shadow AI, introduces direct data privacy risks across corporate networks.

When staff upload internal financial files, customer information, or proprietary source code into public generative models, that data moves outside corporate control and can be stored or used by external providers.

The problem quickly turns into a compliance issue. Under data protection frameworks like the GDPR, routing personal data through unvetted third-party services creates unrecorded processing and potential reporting obligations. Flat bans on AI tools rarely work in practice, as employees routinely find workarounds on personal devices or secondary browsers to keep using them.

Solving shadow AI requires moving from strict blocking to controlled access. Companies need real visibility into endpoint data flows and vetted, private environments where data never leaves corporate boundaries. When teams have compliant, accessible tools for their daily work, the incentive to use unapproved external services disappears.

Does your organization have a clear policy on which AI tools employees can use with company data?

19/10/2026

As commercial trucks, delivery fleets, and rail networks across Europe become increasingly connected, securing onboard systems against digital threats has moved to the top of the agenda for transport teams. From telematics units and diagnostic ports to over-the-air updates, modern vehicle architectures create entry points that demand structured risk management and constant visibility.

Meeting European standards like UNECE R155/R156 and the NIS2 Directive requires clear governance, lifecycle risk assessments, and real-time monitoring via a Vehicle Security Operations Center (vSOC). Check out our full guide to understanding fleet cybersecurity and see how specialized platforms help protect connected transport networks: https://www.vivetech.com/hirek/understanding-fleet-cybersecurity

16/10/2026

The competition among AI model developers is shifting from breaking raw benchmark records to improving operational efficiency. Recent releases such as Claude Sonnet 5.5 and GPT-6.1 Sol show the rise of what many are calling "effort economics": labs are building architectures specifically to cut per-task token costs and reduce latency. Instead of simply scaling parameter counts, the focus has moved to direct, task-based compute allocation.

For enterprise deployments, this changes how teams choose models for production workloads. Organizations are moving away from running the largest, most expensive systems by default. With configurable effort-level settings, teams can match compute expenditure to the actual complexity of a task, cutting costs by 50% to 70% while getting comparable results. Practical ROI now comes from managing token consumption rather than paying for maximum model size.

15/10/2026

As enterprises deploy autonomous AI agents to parse inbound emails, scrape web content, and query corporate databases, a silent threat has emerged: indirect prompt injection.

Unlike direct attacks where a user attempts to jailbreak a chatbot in a prompt window, indirect injection occurs passively when an AI reads third-party content containing hidden instructions. Because Large Language Models process untrusted data within the same context window as system commands, an innocent-looking PDF or promotional email can trick an agent into executing unauthorized tool calls or silently leaking private files.

Traditional firewalls and regex filters fail to catch these malicious payloads because they are written in plain, natural language. Securing connected AI workflows requires an AI-native defense strategy, combining dual LLM architectures to isolate raw external text, real-time semantic guardrails, and mandatory human-in-the-loop controls for high-risk actions. Read our full guide to understand how indirect prompt injection works and how to protect your enterprise AI stack: https://www.vivetech.com/hirek/what-is-indirect-prompt-injection

14/10/2026

Felköszöntöttük a szülinaposainkat! 🎂🎉

A múlt heti kerti partynkon a friss hamburgerek és a Fehér Nyúl sörei mellett a legszebb pillanatot egyértelműen a közös ünneplés jelentette, a videón el is csíptük a gyertyaelfújást!

A nap másik nagy kedvence Mázli, aki egy pillanatra sem tévesztette szem elől a grillezőt (hátha leesik egy kis minőségellenőrzésre váró falat).

Szuper este kerekedett belőle, boldog szülinapot még egyszer az ünnepeltjeinknek!

---

We celebrated our birthday stars! 🎂🎉

At last week’s garden party, we enjoyed fresh burgers, Fehér Nyúl beers, and a special moment together: singing happy birthday and blowing out the candles!

Another star of the day was Mázli, who kept a close eye on the grill, just in case a little bite was available for quality control.

It was a great evening! Happy birthday again to everyone celebrating!

13/10/2026

This Friday, our partner Dyntell Software is hosting their anniversary Dyntell NEXT summit, and the agenda covers some highly relevant themes for today's business leaders.

The event takes a close look at the relationship between emerging technology and human performance. Attendees will hear presentations on enterprise AI and data-driven efficiency, alongside practical insights into executive longevity, stress management, and maintaining peak mental focus.
It promises to be an excellent day for decision-makers gathering in Inárcs to connect and exchange ideas.

Take a look at last year's mood video below for a quick preview, and find the full program details and registration link here: https://dyntellnext.hu/

https://www.youtube.com/watch?v=hcikLZNbs8w&time_continue=0&source_ve_path=NzY3NTg&embeds_referring_euri=https%3A%2F%2Fdyntellnext.hu%2F&embeds_referring_origin=https%3A%2F%2Fdyntellnext.hu

08/10/2026

Viral claims about AI agents rebuilding decades-old software suites overnight make for great headlines, but the reality is more practical. Generative tools cannot magically reconstruct complex binary applications into pristine source code with one click. What they do instead is automate reverse engineering at scale, analyzing compiled structures and generating clean-room functional clones in days rather than months.

This speed changes the threat model for enterprise software and client-side code. Intellectual property built into desktop binaries or distributed applications is much easier to inspect and copy. Threat actors gain the same speed boost, using AI agents to probe binaries for structural flaws, map control flows, and discover zero-day vulnerabilities long before traditional security audits flag them.

Relying on basic binary obfuscation or static packing is no longer enough when analysis runs at machine speed. Protecting proprietary logic requires shifting toward zero-trust application architectures, moving critical operations server-side, enforcing strict runtime integrity, and leveraging secure enclaves. Software creators must build security into the design rather than assuming compiled code remains a safe secret.

How is your organization adapting its application security as automated reverse engineering becomes widely accessible?

Cím

Szüret Utca 15
Budapest
1118

Nyitvatartási idő

Hétfő	09:00 - 17:00
Kedd	09:00 - 17:00
Szerda	09:00 - 17:00
Csütörtök	09:00 - 17:00
Péntek	09:00 - 17:00

Telefonszám

+36 1 209 0196

Weboldal

http://www.vivetech.com/, https://vivesecbox.com/hu, https://www.vivelabergo.com/

Értesítések

Ha szeretnél elsőként tudomást szerezni ViVeTech új bejegyzéseiről és akcióiról, kérjük, engedélyezd, hogy e-mailen keresztül értesítsünk. E-mail címed máshol nem kerül felhasználásra, valamint bármikor leiratkozhatsz levelezési listánkról.