Estimated reading time: 8 minutes
ZeroFox predicts that in 2027, attackers will use help desks, software links, vendor logins, and AI agents to gain illicit access. The company assigns a probability to each scenario. Ransomware is at an all-time high. Many of these forecasts are consistent with what other sources have told CINI this year.
Attackers are more often using access points that companies already trust. Scenarios might include a help desk might reset a password for someone who sounds convincing. A vendor’s login can reach customer data. An AI agent may have ongoing access to company systems.
ZeroFox expects these trends to continue in 2027. Its Threat Forecast for 2027 says attackers “will likely target what organizations already trust.” ZeroFox offers threat intelligence, brand protection, and dark web monitoring. Two of the report’s 13 recommendations suggest subscribing to its services.
Ransomware Forecast: Another Record Year Likely
Ransomware and digital extortion set a record in 2026, ZeroFox says. It counted about 718 incidents a month through October. That compares with about 573 a month in 2025. The third quarter alone brought at least 2,381, the most of any quarter on record.
ZeroFox predicts that 2027 will likely see even more incidents than 2026. The first quarter is expected to start with at least 2,000 cases. The increase is due to more active groups. In 2026, Qilin was responsible for about 15% of tracked incidents, followed by The Gentlemen.
Extortion attacks that do not use encryption will likely make up a bigger share of serious incidents, a trend already underway. These attacks use stolen SaaS data, identity access, and vendor integrations. Fewer victims are expected to pay, but the average payment amount will probably increase.
Other trackers count differently but saw the same climb. GuidePoint Security counted a record 2,760 victims on leak sites in the third quarter. It also found fewer clients paying ransoms, with those who paid paying more. Zscaler’s leak-site count for the year to March 2026 fell 3%. Its window closed before the third-quarter jump.
AI Agents Move Into Routine Attacks
A Russian group linked to the state used AI agents to monitor if defenders detected its malware. When detection happened, the group rebuilt the malware to evade capture, according to Anthropic. ZeroFox sees this as proof that AI is now being used not just to write lures, but also as part of the attack process itself.
ZeroFox says agentic AI will likely shift from experiments to a regular part of intrusion campaigns. Less skilled attackers will benefit the most. The time between a flaw being disclosed and widespread exploitation will probably drop to just hours.
ZeroFox also expects attacks targeting enterprise AI agents to increase faster than other attacks that use AI to create malware. The report notes that having valid agent access can bypass malware detection completely. This forecast is considered likely.
One forecast may already be arriving. ZeroFox rates it likely that a campaign hitting dozens of organizations fast will be publicly tied to AI tools. CINI has reported two that come close. Gambit Security found one operator using AI agent tools to compromise at least 27 online retailers in six days. GreyNoise research cited by GuidePoint showed AI agents compromising at least 11 organizations in 26 seconds.
AI credentials are already being targeted. Bitsight discovered a criminal hacking platform that uses stolen AI keys. The owner of the key ends up paying for the attack.
Social Engineering Moves To Voice And Sessions
The report says social engineering will almost certainly stay the main way attackers get in. File that in: “If it ain’t broke, don’t fix it.” The focus is moving from email to voice calls, collaboration platforms, and help desks. Attackers are also targeting sessions that have already passed multifactor authentication (MFA).
By late June 2026, weekly voice phishing attempts on Microsoft Teams were nearly 10 times higher than the mid-2025 baseline. ZeroFox credits Microsoft for this data.
The report notes that phishing-resistant MFA is causing attackers to focus more on tokens and sessions. It predicts that device-code and consent-phishing attacks will make up a bigger share of identity compromises. CINI has reported on how device-code phishing tricks employees into completing MFA for the attacker.
Robeson Jennings is ZeroFox’s senior vice president of global services and intelligence. Attackers are placing more weight on “manipulating behavior, environments, and outcomes rather than simply breaching systems,” Jennings said.
ZeroFox thinks another multi-victim compromise through a SaaS or third-party integration is very likely in 2027. From August 2025 to June 2026, attackers using the ShinyHunters name stole access tokens from Salesloft Drift, Gainsight, and Klue. These tokens gave access to customers’ Salesforce data. Nudge Security CEO Russell Spitler says he has not yet seen an insurer ask for a company’s SaaS app list.
ZeroFox says real-time voice cloning will probably appear in a high-value incident. Prompt injection attacks on company AI assistants are also likely to cause a public incident involving real data exposure.
Access Brokers Go Private
Initial access brokers sell ways to get into company networks. Their public listings dropped from about 620 in early 2025 to about 370 a year later. ZeroFox links this drop to law enforcement taking down the XSS, RAMP, and BreachForums forums. The company does not see this as a sign that ransomware is slowing down.
Sales are shifting to private channels. Prices for lower-level access dropped from about $1,427 in early 2023 to under $500 by the third quarter of 2026. Domain administrator credentials averaged around $8,000.
ZeroFox expects to see more listings for stolen session tokens and identity-provider access. Access to vendors with customer credentials will likely be more valuable, since one breach can affect many victims. ZeroFox says its 2026 prediction of a more specialized, automated access market was accurate.
Law enforcement is very likely to take down at least one more major cybercrime forum in 2027. However, takedowns have their limits. Lumen told CINI that botnets often rebuild after being taken down.
Geopolitics Will Shape Who Gets Targeted
When it comes to who needs to be most on guard, ZeroFox says companies that develop advanced technology, especially semiconductors and AI systems, will probably face more attacks. This prediction is linked to ongoing competition between the United States and China. Iran is also expected to increase attacks on the Gulf states as well as Turkey and Pakistan. Finally, Russian hybrid activity against European supporters of Ukraine will almost certainly rise, including cyberattacks and influence campaigns targeting European elections in 2027.
Other Forecasts CINI Has Covered
ZeroFox’s outlook is similar to others that CINI has reported on this fall:
- Kroll’s Max Henderson sees no peak yet in ransomware.
- Coalition expects PHP2Shell claims to lag the attacks by weeks, then arrive.
- Allianz Commercial calls confirmed AI losses the tip of the iceberg.
- Gambit Security’s Jake Bernardes says next year is about recovery.
- PwC found boards already judge cyber strategy partly by recovery time and insurability.
ZeroFox recommends starting with an incident response strategy, quick patching, and phishing-resistant MFA. The company will review the forecast in a webinar on Dec. 1.
From The CINI Podcast
ZeroFox maps the threats coming in 2027. Jake Bernardes of Gambit Security says the year will be about recovering from them. With AI, one attacker can now look like a nation-state.
Watch on YouTube | Listen on Spotify | Apple Podcasts | Amazon Music | Read the story
FAQ – 2027 Cyber Threat Forecast
What does ZeroFox’s 2027 cyber threat forecast predict?
ZeroFox expects attackers to target what companies already trust: help desks, SaaS integrations, vendor credentials and AI agents. It also expects ransomware to stay at record levels.
How fast could attackers exploit new flaws in 2027?
ZeroFox says the gap between a flaw’s disclosure and mass exploitation will very likely narrow to hours. Attackers are automating the work with AI agents.
Will ransomware keep rising in 2027?
ZeroFox rates it very likely that 2027 incidents surpass 2026. It counted about 718 incidents a month through October 2026, against about 573 in 2025.
What is extortion without encryption?
Attackers steal data and threaten to leak it without locking any systems. Backups do not stop the threat. ZeroFox expects these attacks to take a larger share of high-impact incidents.
Can attackers get past multifactor authentication?
Yes. ZeroFox expects more attacks that steal session tokens or trick staff into approving access. Those methods work after MFA has been completed.
What do ZeroFox’s probability terms mean?
“Likely” means a 55% to 80% chance. “Very likely” means 80% to 95%. “Almost certain” means 95% to 100%.
Related Cyber Insurance Posts
- AI Losses Are Hiding Inside Other Insurance Claims, Allianz Says
- Cyber Insurance Market Faces Slowdown as SMEs Hold the Key to Future Growth(Opens in a new browser tab)
- Cyber Insurance Market Set to Grow by $42.81 Billion by 2027(Opens in a new browser tab)
- Cyber Insurance Market, Size, Set For Growth(Opens in a new browser tab)
- Global Cyber Insurance Market to Grow to $68.17 Billion by 2030(Opens in a new browser tab)