MITRE ATT&CK Enterprise Round 7 (ER7), published in December 2025, has exposed critical gaps in the cybersecurity industry's ability to defend against modern adversaries. The evaluation, which emulated Scattered Spider and Mustang Panda attack chains, found that the best-performing vendor blocked only 31% of adversarial techniques, with all nine participants achieving 0% protection against identity-based attacks and near-zero cloud attack blocking. Three of the largest vendors—Microsoft, SentinelOne, and Palo Alto Networks—withdrew before the evaluation, citing internal priorities. The results underscore a structural failure in detection-based architectures.
VectorCertain LLC, a relatively small player, took the opposite approach. Using the same MITRE adversary emulations, the company ran its SecureAgent platform through a self-evaluation, blocking 100% of 14,208 tests across three adversaries, including Volt Typhoon. SecureAgent's four-gate governance pipeline evaluates actions before execution, using policy-based controls rather than signatures. This architecture, VectorCertain argues, is why it succeeded where others failed: identity abuse, which relies on legitimate authentication flows, generates no endpoint telemetry, rendering EDR sensors blind.
The implications are significant. Global fraud losses reached $485.6 billion in 2023, and IBM's 2025 Cost of a Data Breach Report pegs the average breach at $4.44 million. IBM found that AI in prevention workflows saved $2.22 million per breach, the largest cost-reduction factor. VectorCertain terms the ongoing losses a "7% Global AI and Cybersecurity Tax"—a compounding cost of architectural failure. The company has enrolled in MITRE's ER8 evaluation, set to introduce composite scoring, and positions SecureAgent as the first AI safety and governance platform to enter the program. The full methodology and results are available at vectorcertain.com.
ER7 data is published at evals.mitre.org.
