Kaspersky SAS 2026 Puts AI Security and Automated Vulnerability Research in Focus

Kaspersky will hold its 18th Security Analyst Summit in Bali from October 20–23, 2026.

Kaspersky will hold its 18th Security Analyst Summit in Bali from October 20–23, 2026.

The agenda includes research into UEFI bootkits, Lazarus-linked activity and attacks against mobile and automotive systems.

The agenda includes research into UEFI bootkits, Lazarus-linked activity and attacks against mobile and automotive systems.

Researchers will examine vulnerabilities affecting gaming platforms, Android TV devices and iOS security.

Researchers will examine vulnerabilities affecting gaming platforms, Android TV devices and iOS security.

AI security sessions will explore attacks against AI coding guardrails and automated vulnerability discovery.

AI security sessions will explore attacks against AI coding guardrails and automated vulnerability discovery.

The SAS CTF 2026 finals will also take place during the event, with an $18,000 prize pool.

The SAS CTF 2026 finals will also take place during the event, with an $18,000 prize pool.

AI Security Takes a Central Role at SAS 2026

Kaspersky is putting attacks against artificial intelligence systems alongside conventional cyberespionage and vulnerability research at its 2026 Security Analyst Summit, reflecting the expanding security surface created by AI-powered software development and automation.

The 18th edition of the Security Analyst Summit (SAS) will take place from October 20 to 23 at the InterContinental Bali Resort in Indonesia. Kaspersky says more than 200 participants from around the world are expected to attend, with the programme featuring vulnerability disclosures and investigations into attacks across multiple technology sectors.

Among the most relevant areas for enterprise technology leaders is the conference's focus on attacks that target AI-assisted development before an AI model has even been invoked.

That research points to a changing cybersecurity environment in which protecting the AI model itself is only one part of the security challenge.

Researchers to Examine Attacks Before AI Is Invoked

One of the most notable SAS 2026 sessions will examine what Kaspersky describes as “pre-prompt attacks.”

Satoki Tsuji of Ikotas Labs is scheduled to demonstrate zero-click remote-code-execution chains that can bypass AI-coding guardrails before an LLM is invoked, according to Kaspersky's published agenda.

The research is relevant to enterprises increasingly deploying AI coding assistants and autonomous development tools.

AI development environments introduce additional components into the software supply chain. Developers may rely on coding agents to interpret instructions, access files, execute commands or interact with development infrastructure.

If an attacker can compromise that environment before the model receives its intended prompt, traditional approaches to AI safety may not be sufficient.

The research therefore highlights a broader security principle: AI applications need to be protected across the surrounding execution environment, not simply at the model or prompt layer.

Automated Vulnerability Discovery Expands the Attack Surface

Another SAS 2026 presentation will explore the use of large language models for automated vulnerability research.

Dr. Zhiniang Peng of Huazhong University is scheduled to present an autonomous, LLM-driven vulnerability-mining system that Kaspersky says discovered and verified more than 100 zero-day vulnerabilities across major Android distributions used by multiple smartphone manufacturers.

The development illustrates the dual-use nature of AI in cybersecurity.

AI can help defenders analyse large codebases and identify weaknesses faster, but the same capabilities can potentially be used to accelerate vulnerability discovery by attackers.

For security teams, this creates pressure to improve the speed of vulnerability management. If automated tools can identify weaknesses at a much faster rate, organisations may need corresponding improvements in patching, validation, software inventory and exposure management.

The presentation is therefore relevant beyond mobile security. It points toward a future in which both offensive and defensive security operations increasingly incorporate autonomous AI systems.

Cyberespionage Research Remains a Major Focus

SAS 2026 will also retain a strong focus on advanced cyberespionage.

Kaspersky GReAT researcher Georgy Kucherin is scheduled to present research into a stealthy UEFI bootkit campaign targeting diplomatic entities, while Sojun Ryu will new findings concerning the Lazarus advanced persistent threat group.

UEFI-level attacks are particularly significant because they operate beneath the conventional operating-system layer.

Compromising the boot process can give attackers a persistence mechanism that is difficult to detect using security controls focused primarily on applications and operating-system activity.

The inclusion of this research alongside AI-focused security demonstrates that the threat landscape remains broad. Enterprises must contend simultaneously with long-established techniques such as firmware compromise and newer attack paths created by AI-enabled development.

Mobile, Automotive and Gaming Systems Under the Microscope

The summit's agenda extends beyond enterprise infrastructure into consumer and connected-device ecosystems.

Kaspersky researcher Dmitry Kalinin is scheduled to present findings on a campaign targeting Android-based automotive head units. The agenda also includes research into security bypasses affecting the PlayStation 5 and a remote-code-execution vulnerability associated with Counter-Strike and Dota 2 that researchers say could provide access to gamers' PCs.

Other sessions will examine the “Kimwolf” Android TV botnet, which Kaspersky describes as a 31.4 Tbps botnet, and “Rocket,” an iOS security bypass associated with the Coruna exploit kit.

These cases demonstrate how the expanding number of connected consumer platforms creates additional opportunities for attackers.

Connected vehicles, smart TVs, mobile devices and gaming systems increasingly combine networking, software and cloud services. Vulnerabilities in any one component can consequently have implications beyond the original device.

The Enterprise Implications of AI-Driven Attacks

For enterprise security teams, the most important takeaway from the SAS 2026 agenda may be the convergence of AI and traditional attack techniques.

AI is becoming part of software development, vulnerability research and security operations, while attackers are simultaneously looking for ways to exploit AI-enabled environments.

That creates several areas organisations will need to monitor.

AI coding assistants should be treated as part of the development environment rather than isolated productivity tools. Their permissions, access to repositories, ability to execute commands and interaction with external systems all become relevant security controls.

Similarly, autonomous vulnerability-discovery systems could change the economics of security research. Organisations may need to identify and remediate weaknesses faster as automated discovery becomes more capable.

The result is a security environment where AI governance, application security, endpoint protection, identity controls and vulnerability management increasingly overlap .

SAS CTF Brings Offensive and Defensive Skills Together

The Security Analyst Summit will also host the finals of the SAS Capture The Flag competition.

Thirteen teams are scheduled to compete in the Attack-Defence format. Eight teams qualified through professional SAS CTF competition, while five regional winners come from the global Kaspersky{CTF} tournament. The event carries an $18,000 prize pool.

The competition is part of the wider research and practitioner focus of the summit, bringing hands-on security testing alongside presentations on advanced threats.

Kaspersky has described SAS as a platform for researchers and security professionals to exchange discoveries and build collaboration across the cybersecurity community.

With AI now appearing directly in the conference's vulnerability research agenda, SAS 2026 provides a snapshot of how cybersecurity research is evolving alongside enterprise adoption of AI.

The key question for security leaders is no longer simply how to secure AI models. It is how to secure the entire environment in which AI agents, coding systems, applications, devices and people interact .

Experienced Associate with a demonstrated history of working in the information technology and services industry.