AI can make parts of an attack faster and easier to scale, and it can help defenders find and fix flaws too. Official sources from NIST and CISA support that two-sided capability picture. They do not measure how much AI has lowered the cost of attacking software, so “cheaper to attack” is best read as a thesis to test rather than an established trend. Defenders can still change the price of a successful attack in their own environments, mainly by removing easy flaws and unnecessary exposure and by fixing exploited vulnerabilities faster.
What the evidence supports and what it does not
Two official sources carry most of the weight. NIST says AI can give defenders new tools to address vulnerabilities, and it can also enhance the capabilities of people targeting organizations and individuals. CISA’s announcement of its vulnerability review for fiscal years 2024 and 2025, dated August 26, 2026, says emerging technology such as AI introduces efficiencies that threat actors can use to automate and scale activity. Those statements support a two-sided capability argument. They do not show that every attack is now cheaper, or that AI alone explains any rise in incidents.
As an Amazon Associate I earn from qualifying purchases.
| Claim | Status | Basis |
|---|---|---|
| AI is dual-use: it can give defenders new tools for vulnerabilities and enhance attacks on organizations and individuals | Supported | NIST’s AI security and resilience page, updated August 14, 2026 |
| AI can introduce efficiencies that threat actors use to automate and scale activity | Supported | CISA’s announcement of its FY2024–FY2025 vulnerability review, August 26, 2026 |
| Attacking software now costs measurably less | Not established | Not stated in the CISA or NIST materials cited here |
| AI-enabled attacks succeed more often than other attacks | Not established | Not stated in the CISA or NIST materials cited here |
| AI alone caused a rise in incidents | Not established | Not stated in the CISA or NIST materials cited here |
Why “cheaper” is a plausible mechanism, not a measured trend
An attack has costs in skilled labor, time spent finding a usable flaw, writing or adapting an exploit, and iterating when a target resists. AI-assisted coding, automated vulnerability discovery, and scripted workflows could reduce some of those costs. That is the mechanism behind the title. The official sources reviewed describe the capability but do not quantify how much it reduces effort or spending for any group of attackers.
The effect is also unlikely to be uniform. This is an inference from how automation works, not a finding from the sources: automation may lower the barrier most for less experienced actors, and it may matter most where targets are reachable and their flaws are already well known.
#1 Best Overall
To move from thesis to established trend, the evidence would need to include:
- Measured effort or cost per successful intrusion, compared across time periods, with the measurement method disclosed.
- The elapsed time between public disclosure of a flaw and its exploitation in the wild, tracked over time.
- The share of known exploited vulnerabilities whose exploitation was automated, with a stated definition of “automated.”
Simple flaws still set the attacker’s entry price
CISA’s review says many actors scan for and exploit simple known vulnerabilities. The recurring weaknesses it highlights are:
- Improper input validation, where software accepts data it should reject.
- Memory safety vulnerabilities, flaws in how software manages memory.
- Poor patching, where known fixes are not applied.
- End-of-support technology, which no longer receives fixes.
If an attacker’s cheapest route is a known flaw in an exposed system, the most direct way to raise that route’s cost is to remove the flaw or the exposure. AI changes how quickly that cheap route can be found; it does not change the basic logic of closing it.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsHow defenders can change the price
In practical terms, the price of an attack is what it costs an adversary to find a working path into your environment and keep using it. Defenders raise that price by reducing the number of easy paths, shortening how long each one stays open, and making the remaining paths harder to use at scale.
Remove common flaws before release
NIST Special Publication 800-218A, finalized July 26, 2024, adds AI-specific practices and tasks to the Secure Software Development Framework (SSDF) 1.1. It is written for AI model producers, AI system producers, and acquirers. Applying it alongside SSDF 1.1 across the model and system lifecycle addresses weaknesses in design and build, rather than leaving them to be found after an incident.
Shrink exposure
Start with an inventory of internet-exposed systems and end-of-support technology, then rank them by exposure and technical impact. CISA points to no-cost Cyber Hygiene scanning and its Internet Exposure Reduction Guidance as starting resources. Every system removed from the internet or retired from support is an entry point that automated scanning can no longer reach.
Prioritize by exploitation risk, not severity alone
CISA lists four prioritization criteria: exposure, known exploited vulnerability (KEV) status, potential for automated exploitation, and technical impact. A workable sequence built on those criteria looks like this:
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →- Build the asset inventory and mark which systems are internet-facing.
- Check each open vulnerability for KEV status, and use the machine-readable signals provided through CISA’s Vulnrichment material.
- Assess whether exploitation of each flagged item could be automated in your environment.
- Remediate the highest-ranked items first.
- Confirm that each fix has taken effect by rescanning or re-testing, rather than assuming it has.
CISA also points to its Stakeholder-Specific Vulnerability Categorization material for aligning these categories with the roles inside your organization.
Best Value
Secure AI deployments, not only AI builds
Organizations that deploy externally developed AI systems face a different job from the teams that build them. CISA’s joint guidance on deploying AI systems securely, announced April 15, 2024, focuses on operating these systems. It addresses confidentiality, integrity, and availability; mitigation of known vulnerabilities; and protection, detection, and response. Builders should work from the development guidance above, and deployers from this operating guidance. The two are complementary.
Measure time, not only activity
CISA’s Cybersecurity Strategic Plan names time to detect adversary activity and time to fix KEVs as measures of effectiveness. Public descriptions of the plan give no numeric target, so set your own baseline and track it over successive quarters. A short fix time matters only if the fix removes real exposure, so pair the metric with a check that the vulnerable path is actually closed.
Quick Recap
Options compared
| Lever | Applies to | Source guidance | What it changes for an attacker |
|---|---|---|---|
| Secure development with AI-specific practices | AI models and systems during build | NIST SP 800-218A with SSDF 1.1, final July 26, 2024 | Fewer exploitable flaws reach deployment |
| Secure AI deployment | Externally developed AI systems in operation | CISA joint guidance, announced April 15, 2024 | Protection, detection, and response for deployed systems |
| Exposure reduction | Internet-facing and end-of-support systems | CISA Cyber Hygiene scanning and Internet Exposure Reduction Guidance | Fewer reachable entry points |
| KEV-first prioritization | Known exploited and automatable vulnerabilities | CISA review criteria, announced August 26, 2026 | Shorter window for flaws already in active use |
| Time-based measurement | Detection and KEV remediation | CISA Cybersecurity Strategic Plan | Shows whether controls reduce real exposure; no numeric target in public descriptions |
Limits of the evidence
- No official source measures how much AI has reduced the cost of attacking software. Treat the title’s cost claim as a thesis.
- NIST describes its AI security work as ongoing and says existing frameworks do not comprehensively address several AI-specific attacks and attack surfaces. Expect the guidance to change.
- CISA’s announcement summarizes its full vulnerability review. For detailed counts and tables, read the full review; this article does not reproduce them.
- Patching is necessary but not sufficient. The sources pair it with prioritization and secure-by-design practices.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
