Google Project Zero is collapsing its public bug-disclosure deadline from 120 days to 30, with the cut rolling in over the next 14 months. That number should rearrange your cybersecurity calendar before a vendor blog does. Thirty days is shorter than most change-advisory boards. It’s shorter than many “critical” patch SLAs, and it’s shorter than the exception ticket you’ve been renewing since last spring. Project Zero is one of the oldest coordinated disclosure programs still running. When it moves the clock, every team that waits on a vendor patch is racing a public write-up, complete with reproduction notes that your threat detection will see about the same time as everyone else’s.
Project Zero Compresses 120-Day Disclosure Into 30 Days
The new rule is simple enough to brief in a standup. A vendor that gets a Project Zero report will have 30 days before the technical details go public, down from the 120-day window that has been the program’s baseline for years. Fourteen months of staged rollout is a courtesy to product teams. Treat it as a warning label for you. If you still brief leadership on quarterly patching, start briefing them on a 30-day public clock for the bugs that will actually get read.
Plenty of operators have watched the old 120-day clock run out while a vendor argued about root cause, affected SKUs, or whether a bug was “exploitable in practice.” Attackers skip those meetings. A shorter deadline makes stalling expensive. It also slams into constraints you already know: certified appliances, medical and industrial software, multi-party firmware, and anything that still needs a maintenance window signed by three managers. Those products will miss day 30. You need a plan that assumes they will.

Researchers are split because both sides are looking at a real failure mode. Disclosure advocates see vendors using four months as slack. Vendor teams see 30 days as a fantasy for anything with a hardware dependency or a customer-controlled upgrade path. Your incident response playbook has to pick a date before the write-up ships. Pick the public date. Extensions will happen. Some will be justified. Building your SLA around an extension you have not been granted is how you end up explaining a weekend scramble to a board that thought “we patch criticals in 30 days” already meant you were fast.
Assume the write-up includes enough to reproduce. That has always been Project Zero’s culture. The new variable is the calendar. Once the notes are indexed, every scanning shop, every exploit broker, and every intern with a lab copy of your edge software gets the same homework. If your cyber security intake still files those bugs into the same 90-day bucket as a CVSS 7.5 on an internal wiki, you will publish your own incident, just later and with worse lighting.
30-Day Cybersecurity Windows Break Standard Patch SLAs
Most enterprises still grade patching by severity and by change window. Critical in 15 or 30 days. High in 60. Everything else in 90, unless someone marks it infeasible, in which case it lives on a risk register that gets a quarterly nod. Project Zero’s clock does not read your GRC spreadsheet. A bug that looks awkward to exploit in isolation can still be hop one once the write-up is public and a second finding is sitting on the host next to it.
The same week as the deadline news, Google described threat actors using AI agents to turn software patches into working exploits. A vendor patch is a diff. A diff is a lesson. If an agent can walk from “fixed in 5.4.2” to a working proof of concept faster than your CAB can meet, then “we are waiting on the patch” is a statement of exposure. You already knew Patch Tuesday is also exploit Tuesday. The new part is the speed, and the lack of a human in the loop for the boring conversion work.
Change control is where this gets ugly. A 30-day public clock does not create a 30-day test environment, a 30-day backup window, or a 30-day vendor for the appliance that last shipped firmware in 2022. You will have systems that cannot take the fix. Document them as standing exceptions with a named isolation owner, a tested rollback, and a date you re-verify the control. An exception without a kill switch is a permission slip for the attacker who reads the same advisory you do.

The underground side of the same compression showed up in an infostealer market Hudson Rock flagged. Stolen logs now ship with AI-written profiles of the people who owned the machines. Buyers search those dossiers in plain language: who works where, which accounts are worth paying for, whether the box looks like a developer laptop or a finance workstation. Credential theft with a product manager attached. Your 90-day SLA is a luxury those buyers do not offer.
This is a bad look for any program that still equates “tracked in the vuln tool” with “handled.” Defense in depth is supposed to cover the gap between disclosure and a clean bill of health. In a lot of shops that gap is a firewall rule from 2019, an EDR policy set to alert-only, and a brute-force watchlist that pages on volume instead of on the one service that still speaks a legacy protocol to the internet. Threat-protection catalogs look complete on a slide. They do not close a 30-day hole on an unpatchable appliance sitting in a plant or a branch office.
Isolation Beats Hope When the Patch Misses Day 30
Start with a list you can finish this week. Which internet-facing products, jump hosts, and crown-jewel adjuncts cannot take a vendor fix in 30 days even if the vendor ships on day 10? Name the owner. Name the compensating control. If the answer is “we’ll see,” you already know how the weekend goes.
Do this immediately, then keep the second half on a schedule:
- Tag every internet-reachable and crown-jewel product that historically misses a 30-day window, and pre-approve an isolation path: firewall deny, service disable, or take-offline.
- On disclosure day, hunt brute-force and exploit scanning against those products first, then rotate standing remote-admin credentials that authenticate to them.
- Split your SLA: in-the-wild, KEV, and Project Zero-class bugs ride a 30-day track with an expiration date on every exception.
- Treat vendor diffs as threat intelligence. Same-day review of what the patch actually changed belongs in detection engineering, not next month’s vuln meeting.
- Make security hardening the default late-patch answer: bind the service to localhost, require a jump host, kill the extra listener, remove the unused feature.
Ongoing work is dull on purpose. Rehearse incident response against a 30-day fuse: vendor notified on day 0, public notes on day 30, exploit chatter on day 31. Who has authority to isolate without a CAB? Who calls the business owner at 2 a.m.? If that conversation currently requires a ticket and a meeting, you will lose the hours that matter. Refresh the isolation runbook when you add a new edge product, not when Google’s tracker does.
Stop letting “infeasible to patch” mean “infeasible to shrink.” An unpatchable box can still lose its WAN path, its extra protocol, and its local admin reuse. That is defense in depth you can ship without a vendor ISO. The teams that get this right will look slow on paper and fast when the write-up lands. That is the trade you want.
Sources
- Google Project Zero Slashes Bug Disclosure Deadlines, Splitting Experts
- Google Finds Hackers Using AI Agents to Build and Run Exploits From Software Patches
- Criminal marketplace sells infostealer logs with AI-written profiles of their owners
Take Control of Your Server Security
Don't let brute-force attacks slow you down. Try IPBan Pro risk-free for 30 days.
Secure. Automated. Lightweight.
Take Control of Your Server Security
Don't let brute-force attacks slow you down. Try IPBan Pro risk-free for 30 days.
Secure. Automated. Lightweight.
