Have we finally solved social engineering? Plus: World Cup fraud, AI IDs and an IBM/OpenAI collab

2026-06-24 · Source: IBM Technology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy, Software Development & Engineering · Depth: Advanced, extended

Summary

A recent discussion explored AI's evolving role in cybersecurity, covering four key areas. First, an op-ed suggests OS-integrated LLMs, like iOS 27's Siri, could significantly reduce social engineering by learning user patterns and providing comprehensive context, similar to how endpoint protection curbed viruses. Second, "Operation Fan Trap" by Cyble Research and Intelligence Labs revealed a massive World Cup fraud ecosystem, identifying nearly 4000 malicious domains exploiting emotional urgency for scams like fake tickets and streaming sites. Third, Estonia is considering granting personal ID codes to AI agents to enhance accountability and permission scoping, addressing the challenge of managing millions of non-human identities. Finally, IBM and OpenAI partnered to launch Security Harness, an application security service leveraging OpenAI's frontier AI models to scan code for vulnerabilities and prove exploits safely within enterprise controls, marking a significant leap in automated vulnerability management.

Key takeaway

For AI Security Engineers developing or deploying agentic systems, recognize that while AI offers powerful defenses against human frailties, it simultaneously introduces new attack surfaces like prompt injection and compromised agent identities. You must prioritize robust identity and access management (IAM) solutions tailored for ephemeral, context-specific agent identities, and integrate advanced telemetry from browser-level sensors to detect novel attack patterns, rather than solely relying on user education.

Key insights

AI presents a dual challenge and solution in cybersecurity, both mitigating and creating new attack vectors.

Principles

• Humans are inherently vulnerable to social engineering due to cognitive overload and lack of context.
• Attackers exploit major events and emotional urgency to industrialize fraud.
• Authenticated malicious agents pose a significant threat, potentially more damaging than unauthenticated ones.

Method

OS-integrated LLMs can provide comprehensive device context to interpret data and flag social engineering attacks. IBM's Security Harness uses frontier AI models to scan application code for vulnerabilities and prove exploits within enterprise guardrails.

In practice

• Implement browser-level sensors to send telemetry to SIEM/EDR for enhanced threat detection.
• Adopt a "guilty until proven innocent" mindset for deals that appear too good to be true.
• Explore short-lived, context-specific tokens for AI agent authentication to mitigate identity theft.

Topics

• Social Engineering
• Large Language Models
• AI Security
• Cybersecurity Fraud
• Identity and Access Management
• Application Security
• AI Agents

Best for: CTO, VP of Engineering/Data, AI Architect, AI Security Engineer, Security Engineer, Director of AI/ML

Related on AIssential

• Can you social engineer an AI? Plus: AI worms and the nonhuman identity problem — AI, despite intelligence, lacks "wisdom" and common sense, making it vulnerable to social engineering and requiring specific guardrails.
• Podcast: Hackers Asked Meta AI To Let Them In. It Worked — The Meta AI incident reveals critical vulnerabilities in AI systems when handling sensitive account management requests.
• The Identity Crisis of AI Agents: Why Autonomous Systems Need IAM Before They Need More… — AI agents require unique, verifiable identities and continuous trust evaluation to prevent an enterprise security crisis.
• Designing AI agents to resist prompt injection — Prompt injection attacks are evolving into social engineering, requiring AI defenses to constrain impact rather than solely filter inputs.
• AI for Criminals — AI significantly amplifies criminal capabilities by providing expertise, precision at scale, and real-time adaptability to defensive measures.
• Is your robot vacuum safe? Here’s why it matters. — Systemic identity failures, unpatched systems, and social engineering remain critical cybersecurity vulnerabilities across industries.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by IBM Technology.