Top news
Investing in the public benefit OfficialAnthropic NewsOct 18Popular Musk Pledges SpaceX Shares to ‘Every Kid in America’ WebBloomberg TechnologyOct 19Popular Websites are closing their doors to AI agents WebThe Verge AIOct 17 This is the new ChatGPT Voice, powered by GPT Live OfficialOpenAI on YouTubeOct 18 How AI giants will reshape global philanthropy WebAxios AIOct 19 Gemini 4 Argon Really Surprised Me WebTwo Minute PapersOct 19 Google Gemini 4 agents just beat Minecraft with no human intervention WebEngadget AIOct 19 Who could it be? 🧐 OfficialGoogle on YouTubeOct 18 Anthropic gives outside AI watchdog “employee-like” access WebArs Technica AIOct 18 What are self-replicating ‘AI worms’, and why is Anthropic worried about them? WebThe Guardian AIOct 18
GlossarySafetyIntermediate

Zero-days

Also written: 0 days, N-days, N-day exploit, Zero-day discovery, Zero-day remote exploits, 0-day remote exploit, Zero-day exploit, Zero-day vulnerability

Unknown software flaws with no available fix, posing significant security risks.

In plain terms

Zero-day refers to a security vulnerability in software that is unknown to the company that made the software. Because it's unknown, there is no patch or fix available for it. This makes it a prime target for attackers who can exploit it before the developers even know it exists.

Imagine a company releases a new app. A hacker discovers a secret way to break into that app, but the company has no idea about this flaw. The hacker can use this 'zero-day' vulnerability to steal data or cause damage until the company eventually finds out and releases a fix.

The expert version

A zero-day vulnerability is a previously unknown flaw in software or hardware for which no patch or fix has been developed by the vendor. Attackers who discover and exploit these vulnerabilities before they are disclosed to the vendor or public gain a significant advantage, as there are no immediate defenses. The term 'zero-day' refers to the fact that the vendor has had zero days to address the vulnerability since its discovery.

Why it matters now

Recent research from Anthropic highlights the potential for advanced AI models to discover and even develop exploits for software vulnerabilities. This raises concerns about AI accelerating the discovery and weaponization of zero-day exploits, impacting cybersecurity.

Where it shows up 13 items, newest first
How each company uses it
Anthropic Claude
Anthropic's research directly investigates how AI models interact with and potentially create exploits for software vulnerabilities, including zero-days.
Related terms

VulnerabilityExploitPatchCybersecurityThreat intelligenceAI safety

Often confused with
N-day exploit
An exploit targeting a known vulnerability for which a patch is a
Vulnerability
A general weakness in software or hardware that could be exploite
Zero-day vulnerability
The flaw itself, whereas a zero-day often refers to the
Origin

The term 'zero-day' originated in the cybersecurity community to describe vulnerabilities and exploits that were unknown to the software vendor.

Try this about 5 minutes

While directly experimenting with zero-days is not feasible or advisable, you can explore AI's role in cybersecurity by asking models like ChatGPT to explain common types of software vulnerabilities. Then, ask it to describe how a hypothetical security patch might fix such an issue.

Try it: Playbook entries that use thisPlaybook →

The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.

Questions people ask

What is the difference between a zero-day and an exploit?
A zero-day vulnerability is the flaw itself, while a zero-day exploit is the code or method used to take advantage of that unknown flaw.
How are zero-days found?
They can be discovered through rigorous security testing, code review, or sometimes by malicious actors.

First seen on Subvolts Jun 9, 2025. Updated Oct 12, 2026. Written by an AI model from the items above and reviewed against them weekly; report an error.