Agentic AI Goes Operational
Agentic AI will very likely move from documented experiments to a routine part of intrusion campaigns in 2027. In 2026, ZeroFox observed AI agents handling reconnaissance, exploitation, and malware rebuilding with limited human direction.
KEY TAKEAWAYS
- The gap between vulnerability disclosure and mass exploitation will very likely shrink to hours as attackers automate the monitor, test, and weaponize cycle.
- Low-skilled actors will very likely show the fastest growth in agentic AI use, operating at a scale once reserved for established groups.
- Attacks targeting enterprise AI agents and their credentials will likely increase more rapidly than attacks using AI to build malware, since valid agent access bypasses malware detection entirely.
- Detection of AI-generated content will very likely remain unreliable in operational environments. Verification through an independent channel will likely remain the most dependable control.
Expect attackers to act before your patch cycle concludes, and limit your AI agents’ access to only what is necessary.

Agentic AI Goes Operational

Geopolitics and Cyber Converge
Geopolitics and Cyber Converge
U.S. efforts to shift security burdens onto allies in Europe, the Middle East, and Asia will likely shape threat actor motivations, capabilities, and targeting in 2027.
KEY TAKEAWAYS
- Russian hybrid measures against European supporters of Ukraine will almost certainly increase, including sabotage, cyberattacks, influence operations, and small-scale drone incursions.
- Russian-backed cyber-influence campaigns will likely target Europe’s 2027 elections, with France leading the cycle.
- Iran will likely intensify targeting of the Gulf states, Turkey, and Pakistan as they take on more of the U.S. security role in the region.
- Threat actors will likely target companies producing advanced technology that competes with China’s interests, especially semiconductors and AI systems.
Prioritize coverage of flashpoints that are most likely to affect your offices, suppliers, and markets.
The Dark Web Goes Private
The deep and dark web (DDW) will almost certainly remain the commercial center of cybercrime in 2027, but trust in public forums is collapsing. Law enforcement seized or compromised the RAMP and LeakBase forums, and XSS and BreachForums suffered disruptions over 2025 and early 2026. ZeroFox expects brand-sharing collectives, invitation-only venues, and private messaging channels to replace forums as the default.
KEY TAKEAWAYS
- Law enforcement will very likely seize or compromise at least one more major cybercrime forum or marketplace in 2027.
- New forums and marketplaces will very likely launch and fail faster than in any prior year, with vouching and invitation-only access deciding who gets in.
- Telegram will likely remain a primary coordination channel, but its share of high-value criminal activity will likely fall as actors move to other encrypted or self-hosted platforms.
- Unverified or exaggerated leak claims will likely become routine, forcing victims and researchers to validate stolen data before acting.
Public chatter will tell you less about your exposure in 2027. Validate every leak claim before you act on it.

The Dark Web Goes Private

Ransomware Breaks Its Own Record
Ransomware Breaks Its Own Record
Ransomware and digital extortion (R&DE) will very likely stay at record levels in 2027. ZeroFox identified more victims in 2026 than in any prior year and assesses that 2027 will likely top it. A rising number of active groups, not one dominant operation, will very likely drive the growth.
KEY TAKEAWAYS
- ZeroFox identified an average of approximately 718 R&DE incidents per month through October 2026, compared to approximately 573 per month in 2025, which was already a record year.
- ZeroFox observed at least 2,381 R&DE incidents in Q3 2026, the highest quarterly total on record. Q1 2027 will likely begin at or above 2,000 incidents.
- Extortion without encryption, built on stolen SaaS data, identity access, and vendor integrations, will very likely account for a larger share of high-impact incidents.
- Manufacturing organizations will very likely face the biggest threat in 2027, with professional services, construction, technology, and healthcare also prone to significant targeting.
Payment rates will likely stay low in 2027, but the victims who pay will likely pay more.

Access Brokers Go Dark
Access Brokers Go Dark
Initial access brokers (IABs) will very likely remain essential to cybercrime in 2027 while getting harder to observe. ZeroFox observed about 370 network access listings on the DDW in Q1 2026, down from 620 in Q1 2025. The drop very likely does not signal less ransomware. ZeroFox assesses it likely reflects brokers moving sales into private channels after law enforcement took down the forums where they advertised.
KEY TAKEAWAYS
- Average prices for lower-tier access fell from about USD 1,427 in early 2023 to below USD 500 by Q3 2026, while domain administrator credentials averaged about USD 8,000.
- Stolen session tokens and SaaS or identity-provider access will very likely appear in listings more often than in 2026.
- The time between compromise and sale will very likely shrink as brokers automate validation and listing.
- Access to vendors and integration providers that hold downstream credentials will likely command a premium, because one compromise can reach many victims.
Watch for your credentials and vendor-held access in the private channels where brokers now sell, not only in public listings.
Live Webinar
Inside the Channels You Can’t See: ZeroFox Analysts on 2027
Tuesday, December 1
ZeroFox Intelligence analysts walk through the 2027 forecasts and what security leaders should plan for first. Join the Americas or EMEA session.
Speakers: Adam Darrah, Craig Lowrie, and William Byers
Sessions: 11 AM ET | 1 PM GMT



Social Engineering Moves Past the Inbox
Social engineering will almost certainly remain the leading way into enterprise environments in 2027. Attackers are moving from email to voice, collaboration platforms, help desks, and sessions that have already passed multifactor authentication (MFA).
KEY TAKEAWAYS
Attackers want sessions that already passed MFA. Verify through an independent channel before your help desk resets any credential.
Social Engineering Moves Past the Inbox