On October 28, 2026, a16z published "The New Security Agenda," a survey of 100 chief information security officers at companies with more than 1,000 employees. Eighty-seven percent expect cybersecurity spending to rise in 2027, nearly half expect their security teams to grow, and only 4% expect cuts. Money and headcount are both moving up, but the most interesting part of the survey is the gap between rising budgets and actual deployment.
A changing agenda: all of the top seven 2027 items involve AI
In respondents' recollection of 2026 priorities, exposure and vulnerability management and SIEM topped the list. For 2027, all seven of the most-selected initiatives involve AI. Internal AI agent security, SOC automation, and offensive security gained the most ground, followed closely by code security, remediation, and AI governance. Nearly half of respondents put AI code security among their top five priorities for 2027, up from about 25% in 2026.
Plenty of pilots, little production use
Only 18% of companies have AI SOC tools in production, and 13% have put AI offensive security or AI exposure and vulnerability remediation into production. Buying plans are moving faster: 45% plan a new purchase or upgrade for internal AI agent security within twelve months, 42% for offensive security, and 38% for AI usage and governance. Eighty-three percent of respondents say more frequent security posture testing matters, yet today only 18% say most offensive testing runs continuously against production assets, while 52% expect to get there within a year. Finding problems faster is not the same as fixing them faster: 70% want retesting to verify that vulnerabilities were actually fixed, 67% want remediation prioritized by validated exploitability, and 57% want fixes or remediation tickets generated automatically.
Incumbents get the first call; startups must fill a gap
Forty-four percent of respondents look first to new capabilities from their existing security platform, and only 21% start with an AI-native vendor; 47% say they would choose a startup only when it solves a capability gap their incumbent cannot address. The other side of the picture is the attack surface itself: 92% of companies already use employee generative AI tools in production, 83% use coding assistants or agents, and 55% have MCP servers or agent integrations in production. Forty-seven percent worry most about employees leaking sensitive data into AI tools, 46% about agents exceeding their permissions or being compromised, and 63% say security reviews have already slowed at least some AI adoption.
This survey should be read with one caveat in mind: a16z is an investor, several companies named in the piece belong to its portfolio, and the figures come from 100 respondents describing themselves and recalling 2026, not from a census of the whole industry. Its value lies in direction: the fight over security budgets has moved from whether to buy AI security tools to who can prove that things actually got fixed. For buyers, next year's selection question is not a feature checklist, but whether a vendor can run the full loop of find, fix, and retest inside their own environment.
