A new class of cyberattack is emerging that does not wait for human operators. It writes its own code, moves laterally through networks, and exfiltrates data — all in under 22 seconds. This is the threat IBM addressed head-on this week, announcing a pair of enterprise cybersecurity products designed to meet agentic AI attacks with agentic AI defenses.
The April 15 announcement introduced two distinct offerings: a frontier model threat assessment delivered through IBM Consulting, and IBM Autonomous Security — a multi-agent service the company describes as capable of coordinating detection, response, and threat intelligence at machine speed.
The 22-Second Problem
The urgency behind IBM’s announcement reflects a genuine shift in the threat environment. Security researchers have documented agentic attack chains in 2026 that collapse what used to be multi-day intrusion timelines into seconds. CyberStrikeAI, one such AI-assisted offensive tool, executed fully automated credential harvesting and network reconnaissance against FortiGate infrastructure across 600+ firewalls in 55 countries. Another campaign, tracked by Microsoft Security, used generative AI to craft targeted phishing emails tailored to victim roles — RFPs, invoices, manufacturing workflows — and automated the full delivery chain.
The Mercor breach, confirmed in early April, added a supply-chain dimension to the threat picture: a $10 billion AI training data startup was compromised through LiteLLM, a component used across the industry, with customers including Anthropic, OpenAI, and Meta potentially affected.
Against this backdrop, traditional human-in-the-loop security operations are structurally disadvantaged. IBM’s bet is that defense must operate at the same speed and autonomy as offense.
What IBM Autonomous Security Actually Does
The new service brings together interoperable, vendor-agnostic digital workers — IBM’s term for specialized security agents — that operate across an organization’s full security stack. The agents coordinate to analyze software exposures and runtime environments, map exploit paths, enforce security policies, detect anomalies, and contain threats. The design goal is minimal human intervention for routine threat response, with human oversight preserved for high-stakes decisions.
The companion enterprise assessment, delivered by IBM Consulting alongside technology partners, audits readiness for agentic-enabled threats specifically: security gaps, policy weaknesses, AI-specific exposures, and exploit paths that traditional assessments may not surface. It delivers prioritized mitigation guidance, including interim safeguards where no immediate software fix is available.
IBM frames both products as architecture-agnostic — a significant selling point for large enterprises running heterogeneous environments — and positions the assessment as a starting point before committing to the managed service.
A Market Under Pressure
IBM’s move is part of a broader industry response to the weaponization of frontier AI models. Google, Microsoft, and Palo Alto Networks have all expanded their AI-native security offerings in the first quarter of 2026, but IBM’s framing of autonomous defense agents as a direct mirror to autonomous attack agents is among the more direct competitive postures taken to date.
The market pressure is real. The inadvertent leak of Anthropic’s Mythos model last week erased $14.5 billion in cybersecurity stock value in a single session, reflecting investor anxiety that powerful, unconstrained models could accelerate attacker capability faster than defenders can adapt.
IBM Autonomous Security does not solve that structural problem — no single product does — but it represents a credible attempt to operationalize defense at machine speed. Whether enterprises adopt it at scale will depend on integration complexity, pricing, and how quickly the autonomous threat landscape continues to evolve. Given the pace of 2026 so far, that evolution shows no sign of slowing.
