The modern security operations center (SOC) is navigating a new era where bad actors use automation relentlessly and continuously adapt tactics. At the same time, the software ecosystem is changing faster than traditional security practices can keep pace.
The rapid adoption of AI-generated code, autonomous systems, cloud-native architectures and increasingly complex software supply chains are fundamentally reshaping the risk landscape. AI is not making development safer. According to Veracode’s Spring 2026 GenAI Code Security Update, AI-generated code fails security checks roughly half the time across the four most critical OWASP vulnerability categories.
Cybersecurity professionals can no longer afford to focus solely on hunting isolated and individual vulnerabilities in production; they must anticipate how software development, deployment and operations are continuously evolving. Staying ahead of these trends is essential to defend critical data, enable rapid innovation and maintain trust.
AI-Generated Code Requires New Validation
Generative AI coding assistants have sparked a massive productivity boom, drastically increasing the volume of code pushed into enterprise environments. But as AI models get better at writing code that looks correct, it continues to introduce the same classes of security flaws. SQL injection, cross-site scripting, log injection and insecure cryptographic algorithms remain as prevalent in AI output today as when testing began. This enormous yield of code can easily overwhelm traditional manual review processes.
While conventional static code analysis tools remain valuable for automating alerts on legacy problems like buffer overflows, they are no longer sufficient. Even after clean static scans, AI-generated code can introduce unintended functionality that passes conventional checks but behaves in risky ways.
By applying testing strategies commonly used in malware analysis, software security professionals can identify when an AI assistant produces code that deviates from its intended purpose.
Dive Deeper: Security Congress Session
- Is Your AI Code Assistant Generating Badly Behaving Code? Strategic Testing Can Help Decide
Monday, October 26: 10:10 a.m. – 11:10 a.m.
Humans and AI Must Become Partners
With AI increasingly integrated into writing and reviewing code, organizations must understand where humans outperform AI and where AI offers distinct advantages. Do AI tools introduce novel vulnerabilities, or do they catch things humans routinely miss?
A recent study of confirmed common vulnerabilities and exposures (CVEs) in a major open-source healthcare platform deployed across 6,000 facilities mapped this exact dynamic. Researchers traced each vulnerability back to the pull request that introduced it, mapping findings into four quadrants based on the OWASP Top 10: what humans introduce, what AI introduces, what humans catch and what AI catches.
The findings show that AI may identify sweeping patterns that human reviewers overlook, while humans provide vital contextual judgment and business awareness. Security programs must develop a collaborative human-AI security review model, intelligently routing vulnerability findings to the reviewer best equipped to resolve them.
Dive Deeper: Security Congress Session- Who Writes the Bug, Who Finds the Bug? Human-AI Security Lessons from Open Source
Tuesday, October 27: 3:00 p.m. – 4:00 p.m.
Front-Line Defense Is a Priority
Software supply chain security has moved from theoretical concern to operational emergency. Security teams increasingly rely on third-party open-source packages, but they lack visibility into transitive dependencies and runtime exposure. Sonatype’s 2026 State of the Software Supply Chain Report catalogued more than 454,600 new malicious open-source packages in 2025 alone, bringing the cumulative total of known and blocked malware to more than 1.2 million packages across npm, PyPI, Maven Central, NuGet and Hugging Face. When a malicious package is indistinguishable from legitimate software, incident response gets tested in production.
Real-world compromises involving npm-supply-chain-attack and sha1-hulud highlight this exact challenge. In these large-scale SaaS environments, runtime triage cannot wait for a clean software bill of materials (SBOM), ownership of transitive dependencies is undefined and attribution hits a wall despite repository scraping and telemetry.
Executable supply chain defense requires runtime visibility across live services, containment procedures that can move faster than an autonomous spreading event and governance that clarifies dependency ownership before an incident forces the question.
Dive Deeper: Security Congress Session
- When the Supply Chain Bites Back: npm-supply-chain-attack, sha1-hulud and Why the Old Playbook Breaks in an AI-Native World
Monday, October 26: 11:20 a.m. – 12:20 p.m.
Identity as the Control Plane for Supply Chains
While the industry often fixates on complex code manipulation, many supply chain compromises originate from identity weaknesses. Overprivileged service accounts, compromised identities and poorly governed access across CI/CD pipelines create massive opportunities for lateral movement and privilege escalation.
Securing every stage of the software life cycle, including source control, build environments, artifact repositories and deployment platforms, demands a fundamental reframing of supply chain security as an identity governance challenge. Defenders must operationalize identity-centric strategies, including strict least privilege enforcement in pipelines, segmentation of machine identities and federated governance models. Integrating identity and access management (IAM), DevSecOps and enterprise risk management reduces the attack surface without slowing development velocity. Future software security programs will increasingly treat identity as their core security layer.
Dive Deeper: Security Congress Session
- Identity as the Control Plane for Software Supply Chain Security
Wednesday, October 28: 10:45 a.m. – 11:45 a.m.
Managing Non-Human Identities
Modern infrastructure runs on machine-to-machine trust. Enterprise environments are powered by an increasingly complex and large workforce of non-human identities (NHIs): service accounts, API tokens, workload identities and CI/CD automation runners. Machine identities now outnumber human identities in the average enterprise by 109 to 1, according to Palo Alto Networks' 2026 Identity Security Landscape report.
Because these non-human identities frequently outnumber human users, they have become a preferred attack vector in supply-chain breaches. Misconfigured automation creates immediate opportunities for privilege escalation and persistence, exposing tokens within minutes. Through frameworks like the NHI Pipeline Attack Kill Chain, defenders can map exactly how machine identities are discovered, stolen, escalated and exploited.
Organizations must build aggressive governance for machine-driven environments. This includes complete life cycle management, rapid detection for compromised credentials and specialized incident response playbooks for automation accounts. Effective governance of this invisible workforce will soon become a defining security capability.
Dive Deeper: Security Congress Session
- The Invisible Workforce: How Non-Human Identities Are Breached and How to Defend and Govern Them
Tuesday, October 27: 4:45 p.m. – 5:45 p.m.
Agentic AI: New Authentication Challenges
As organizations rapidly adopt agentic AI ecosystems, traditional identity models are stretched to their breaking point. These platforms launch short-lived, autonomous agents that act on behalf of users with delegated authority. Existing authentication mechanisms were designed for users and long-lived services, not highly dynamic, ephemeral agents, creating massive gaps in trust, delegation and auditability.
Security leaders must prepare for an agent-driven future by analyzing modern identity protocols, asking themselves how agents should be authenticated and authority be delegated. Emerging frameworks evaluate protocols like SPIFFE for workload identity, OAuth and OpenID Connect for delegation, Token Exchange, DPoP for sender-constrained tokens, CIBA for human step-up and decentralized identity. By understanding such models, organizations can position themselves to handle federation automation and high-churn performance.
Dive Deeper: Security Congress Session
- Bright Ideas Roundtable: A Conceptual Framework for Authentication in Agentic AI Ecosystems: Protocol Analysis and Taxonomy
Monday, October 26: 2:45 p.m. – 3:45 p.m.
AI Transforms DevSecOps and API Defenses
AI is fundamentally transforming both offense and defense within the DevSecOps pipeline. AI-powered penetration testing introduces new ways to identify weaknesses at scale, automating attack simulations and accelerating the discovery of vulnerabilities. Security validation is becoming more continuous and adaptive.
Nowhere is this shift more critical than in API security. Traditional API protections, such as static rate limiting, WAF rules, and OAuth controls, were designed for predictable, noisy attackers. Today, AI-enabled attackers operate with precision, conducting intelligent reconnaissance, staying under throttling thresholds and exploiting open API specification documentation to generate targeted payloads that legacy fuzzing misses entirely. Security teams must collaborate closely to implement anomaly detection, behavioral token analysis and advanced intelligence before attackers exploit these gaps.
Dive Deeper: Security Congress Sessions
- AI Pentesting: Reimagining the DevSecOps Stack
Monday, October 26: 4:30 p.m. – 5:30 p.m. - Tabletop/Gamified: From Fuzzing to Artificial Intelligence: How API Gateway Defenses Must Evolve
Tuesday, October 27: 4:45 p.m. – 5:45 p.m.
Cloud-Native Environments: Uncovering Hidden Risks
Security challenges extend far beyond the application layer. Cloud-native environments have introduced insidious infrastructure-level risks to the mix. As containerized workloads share underlying components, like kernel slab allocators, standard isolation assumptions are becoming increasingly dangerous.
Attackers are actively targeting these shared resources through sophisticated side-channel attacks. The HeapSpy attack demonstrated how adversaries exploit kernel memory allocator timing to infer highly sensitive operations across container boundaries. By sending network traffic triggering socket buffer allocations and measuring syscall latency variations, attackers successfully fingerprinted cryptographic operations (distinguishing AES-128 from AES-256 with 87% accuracy) and detected TLS handshakes in co-located Docker containers.
Strengthening container security requires enhanced workload isolation, namespace hardening, continuous monitoring and vital kernel-level protections like per-cgroup slab isolation and timing jitter to definitively eliminate cross-container data leakage.
Dive Deeper: Security Congress Session
- HeapSpy: Breaking Container Isolation Through Kernel Memory Allocator Side Channels
Tuesday, October 27: 9:25 a.m. – 10:25 a.m.
The Future Belongs to the Adaptable
A common operational thread weaves through these challenges, involving AI-generated software, intelligent attackers, supply chain vulnerabilities, machine identity governance and cloud-native infrastructure risks. Security controls that rely on yesterday's static assumptions will fail under the weight of automated, real-time threats.
The most resilient security programs treat exposure as continuous, governing machine identities with the rigor once reserved for human accounts and build cross-functional coordination as an intentional capability. Software security is no longer a specialty practice detached from development; it is a design discipline running through every layer of how modern software is built, deployed and operated.
The organizations that succeed will be those whose security professionals continuously evolve alongside the software they are tasked with protecting. Staying ahead of what’s next is no longer a competitive advantage but a professional necessity.
