Explainer
Zero-day vulnerabilities explained
A zero-day is a software or hardware flaw that attackers exploit before a fix is publicly available.[1][2] Google's threat intelligence group counted 90 zero-days exploited in the wild in 2025, nearly half of them in enterprise software and appliances, and for the first time attributed more to commercial surveillance vendors than to state espionage groups.[3][4][5]
What makes a vulnerability a “zero-day”
A zero-day is a security flaw that attackers exploit before the maker has released a fix. NIST defines a zero-day attack as one that exploits a previously unknown hardware, firmware or software vulnerability.[1]
Definitions vary at the edges. Google Threat Intelligence Group (GTIG), which publishes an annual review of exploited zero-days, defines a zero-day as a vulnerability maliciously exploited in the wild before a patch was publicly available.[2] Publicly disclosed vulnerabilities are catalogued by the CVE program, one record per flaw; see the CVE program.[6]
How many are exploited
GTIG tracked 90 zero-days exploited in the wild in 2025. That was below the 2023 record of 100 but above 78 in 2024.[3] GTIG cautions that its figures cover only zero-days that were detected and disclosed, and that past counts can change as forensic investigations uncover older incidents.[7]
The targets are shifting toward business technology. GTIG identified 43 zero-days (48%) in enterprise software and appliances in 2025, up from 36 (46%) in 2024.[4] By type, memory safety bugs, especially use-after-free and out-of-bounds write errors, made up roughly 35%.[8]
Who uses them
In 2025, for the first time since it began tracking, GTIG attributed more zero-days to commercial surveillance vendors than to traditional state-sponsored espionage groups. GTIG said this shows these vendors are widening access to zero-day exploitation to more customers than ever.[5] It attributed nine to confirmed or likely financially motivated criminal groups.[9] Internet-facing vulnerabilities are among the entry routes CISA lists for ransomware attacks; see how ransomware works.[10]
Disclosure deadlines
Some researchers who report flaws to vendors set a deadline before going public. Google’s Project Zero, for example, gives vendors 90 days after notification to release a patch and publishes technical details 30 days after the patch is out.[11] CVE IDs can be reserved during this period, so the people involved can coordinate before anything is public.[12]
Once a flaw is known to be exploited, CISA can add it to its Known Exploited Vulnerabilities catalog, which requires reliable evidence of exploitation, a CVE ID and clear remediation guidance.[13] Under Binding Operational Directive 22-01, US federal agencies generally have two weeks to fix catalogued flaws whose CVE IDs were assigned from 2021 onward, and six months for older ones.[14]
Should governments keep zero-days?
Intelligence and police agencies sometimes find flaws they could use themselves. Both the US and UK have formal processes for deciding.
- United States. The Vulnerabilities Equities Process charter of 15 November 2017 says its primary focus is disclosing flaws the government discovers, absent a “demonstrable, overriding interest” in using them for lawful intelligence, law enforcement or national security purposes.[15] An interagency Equities Review Board meets monthly to decide.[16]
- United Kingdom. GCHQ says its Equities Process weighs whether to release a flaw for fixing or retain it for intelligence use, and that the starting position is always that disclosure is in the national interest.[17]
AI systems now find previously unknown flaws as well: in November 2024 Google said its Big Sleep agent had found an exploitable memory-safety bug in SQLite, which it called the first public example of an AI agent doing so in widely used software.[18] For the debate over whether that helps attackers or defenders more, see the AI cybersecurity debate. For CISA’s role, see cisa.
Read next
CISA (Cybersecurity and Infrastructure Security Agency)
The US agency that leads national cyber and infrastructure defense: its KEV catalog, advisories, budget and staff cuts, and the 2015 info-sharing law.
Questions readers ask
What is a zero-day?
NIST defines a zero-day attack as one that exploits a previously unknown vulnerability. Google's threat intelligence group counts a flaw as a zero-day if it was maliciously exploited before a patch was publicly available.[1][2]
How many zero-days are exploited each year?
Google Threat Intelligence Group tracked 90 in 2025, 78 in 2024 and a record 100 in 2023. It notes that counts cover only zero-days that were detected and disclosed.[3][7]
Who uses zero-day exploits?
In 2025 Google attributed more zero-days to commercial surveillance vendors than to traditional state-sponsored espionage groups for the first time, and nine to financially motivated criminals.[5][9]
How long do vendors get to fix a reported flaw?
It depends on who reports it. Google Project Zero, for example, gives vendors 90 days to release a patch and publishes details 30 days after the patch.[11]
Sources
· 18 claims from 8 publishers
Each numbered claim is a statement we checked against the sources listed with it. Tap a number in the text to see its sources in place.
- [1]
The NIST glossary defines a zero-day attack as one that exploits a previously unknown hardware, firmware or software vulnerability. confirmedas of 2026-11-08
- zero day attack (glossary) · NIST CSRC · Primary source · Definitions (CNSSI 4009-2022) (retrieved 2026-11-08)
- [2]
Google Threat Intelligence Group defines a zero-day as a vulnerability maliciously exploited in the wild before a patch was publicly available. confirmedas of 2026-11-08
- Hello 0-Days, My Old Friend: A 2024 Zero-Day Exploitation Analysis · Google Threat Intelligence Group · Primary source · Executive Summary notes (retrieved 2026-11-08)
- [3]
Google Threat Intelligence Group tracked 90 zero-day vulnerabilities exploited in the wild in 2025, below the 2023 record of 100 and above 78 in 2024. confirmedas of 2026-11-08
- Look What You Made Us Patch: 2025 Zero-Days in Review · Google Threat Intelligence Group · Primary source · Executive Summary (published March 2026) (retrieved 2026-11-08)
- [5]
For the first time since it began tracking, GTIG attributed more 2025 zero-days to commercial surveillance vendors than to traditional state-sponsored espionage groups, which it said shows these vendors widening access to zero-day exploitation to more customers. confirmedas of 2026-11-08
- Look What You Made Us Patch: 2025 Zero-Days in Review · Google Threat Intelligence Group · Primary source · Executive Summary (retrieved 2026-11-08)
- Look What You Made Us Patch: 2025 Zero-Days in Review · Google Threat Intelligence Group · Primary source · Executive Summary (retrieved 2026-11-08)
- [6]
The CVE Program says its mission is to identify, define and catalog publicly disclosed cybersecurity vulnerabilities, with one CVE Record for each vulnerability. confirmedas of 2026-11-08
- About the CVE Program (cve.org Overview page source) · CVE Program (CVEProject on GitHub) · Primary source · About the CVE Program (cve.org Overview page source) (retrieved 2026-11-08)
- [7]
GTIG says its counts are based on detected and disclosed zero-days and may change as past incidents are discovered through forensic investigations. confirmedas of 2026-11-08
- Look What You Made Us Patch: 2025 Zero-Days in Review · Google Threat Intelligence Group · Primary source · Scope and definitions (retrieved 2026-11-08)
- Hello 0-Days, My Old Friend: A 2024 Zero-Day Exploitation Analysis · Google Threat Intelligence Group · Primary source · Scope and definitions (retrieved 2026-11-08)
- [8]
GTIG says memory safety issues, particularly use-after-free and out-of-bounds write bugs, accounted for roughly 35% of the zero-days it tracked in 2025. confirmedas of 2026-11-08
- Look What You Made Us Patch: 2025 Zero-Days in Review · Google Threat Intelligence Group · Primary source · Vulnerability types, Memory Corruption (retrieved 2026-11-08)
- [9]
GTIG attributed the exploitation of nine zero-days in 2025 to confirmed or likely financially motivated groups. confirmedas of 2026-11-08
- Look What You Made Us Patch: 2025 Zero-Days in Review · Google Threat Intelligence Group · Primary source · Executive Summary (retrieved 2026-11-08)
- [10]
The #StopRansomware Guide groups its prevention advice by six common initial access vectors: internet-facing vulnerabilities and misconfigurations, compromised credentials, phishing, precursor malware infection, advanced forms of social engineering, and third parties and managed service providers. confirmedas of 2026-11-07
- #StopRansomware Guide · CISA, MS-ISAC, NSA and FBI · Primary source · Part 1, section headings (other vectors are later headings) (retrieved 2026-11-07)
- [11]
Google Project Zero follows a 90+30 policy, giving vendors 90 days after notification to release a patch and publishing details 30 days after a patch is available. confirmedas of 2026-11-08
- Vulnerability Disclosure Policy · Google Project Zero · Primary source · 90+30 policy (retrieved 2026-11-08)
- Vulnerability Disclosure Policy · Google Project Zero · Primary source · 90+30 policy (retrieved 2026-11-08)
- [12]
The CVE Program says a Reserved ID lets stakeholders coordinate on a vulnerability early while the responsible CNA is not yet ready to disclose it publicly. confirmedas of 2026-11-08
- Process: CVE Record Lifecycle (cve.org Process page source) · CVE Program (CVEProject on GitHub) · Primary source · CVE Record Lifecycle, step 4 (retrieved 2026-11-08)
- [13]
To add a vulnerability to the KEV catalog, CISA requires reliable evidence of exploitation in the wild, an assigned CVE ID and clear remediation guidance. confirmedas of 2026-11-07
- BOD 22-01: Reducing the Significant Risk of Known Exploited Vulnerabilities · CISA · Primary source · 2021-11-03 (retrieved 2026-11-07)
- [14]
Under BOD 22-01, agencies must remediate catalogued vulnerabilities within six months for CVE IDs assigned before 2021 and within two weeks for all others, unless CISA adjusts the timeline for grave risk. confirmedas of 2026-11-07
- BOD 22-01: Reducing the Significant Risk of Known Exploited Vulnerabilities · CISA · Primary source · 2021-11-03 (retrieved 2026-11-07)
- [15]
The US Vulnerabilities Equities Process charter of 15 November 2017 says the policy's primary focus is disclosing vulnerabilities the government discovers, absent a demonstrable, overriding interest in using them for lawful intelligence, law enforcement or national security purposes. confirmedas of 2026-11-08
- Vulnerabilities Equities Policy and Process for the United States Government · The White House (archived) · Primary source · Title page (retrieved 2026-11-08)
- Vulnerabilities Equities Policy and Process for the United States Government · The White House (archived) · Primary source · 1. Purpose (retrieved 2026-11-08)
- Vulnerabilities Equities Policy and Process for the United States Government · The White House (archived) · Primary source · 1. Purpose (retrieved 2026-11-08)
- [16]
Under the 2017 charter, an interagency Equities Review Board is the main forum for VEP decisions and meets monthly. confirmedas of 2026-11-08
- Vulnerabilities Equities Policy and Process for the United States Government · The White House (archived) · Primary source · 4.1 Equities Review Board and VEP Director (retrieved 2026-11-08)
- [17]
GCHQ says its Equities Process decides whether vulnerabilities it finds are released for fixing or retained for intelligence use, and that the starting position is always that disclosure is in the national interest. confirmedas of 2026-11-08
- The Equities Process · GCHQ · Primary source · The Equities Process (retrieved 2026-11-08)
- The Equities Process · GCHQ · Primary source · The Equities Process (retrieved 2026-11-08)
- [18]
In November 2024 Google's Big Sleep AI agent was reported to have found a previously unknown exploitable memory-safety bug in SQLite, which Google called the first public example of an AI agent doing so in widely used real-world software. confirmedas of 2024-11-01
- From Naptime to Big Sleep: Using Large Language Models To Catch Vulnerabilities In Real-World Code · Google Project Zero · Primary source · 2024-11-01 (retrieved 2026-10-10)
Revision history (1)
- Page created.
Created Nov 8, 2026. Last reviewed by an editor on Nov 8, 2026. Next scheduled review: Mar 8, 2027.
Cite this page
"Zero-day vulnerabilities explained." ContentLora, updated Nov 8, 2026. https://contentlora.com/explain/zero-day-vulnerabilities-explained
Spotted an error? Suggest a correction or emailcorrections@contentlora.com.