📊 Full opportunity report: The 90-Day Window Closed. Nobody Sent a Notice. on ThorstenMeyerAI.com — validation score, market gap, and execution plan.

TL;DR

The 90-day coordinated disclosure period has expired without any vendors issuing notices or patches. This shift in vulnerability management raises questions about the future of responsible disclosure and security response times.

The 90-day coordinated disclosure window, a cornerstone of cybersecurity practices since the early 2000s, has officially closed without any notices or patches from affected vendors, marking a significant shift in vulnerability management.

On April 1, 2026, a critical Linux kernel vulnerability known as Copy Fail was patched in the mainline kernel. Despite the patch being publicly available from that date, no vendor or affected organization issued a security notice or deployed a patch within the subsequent 28 days, which ended on April 29, 2026.

This absence of notices is notable because the traditional 90-day window was designed to give defenders time to respond while limiting attacker opportunities. The delay or absence of vendor disclosures suggests a fundamental change in how vulnerabilities are managed, especially with AI-driven tools capable of rapidly analyzing patches and developing exploits.

Security researchers and industry observers are now questioning whether the responsible disclosure model remains effective in its current form, given the accelerated pace of vulnerability discovery and exploit development facilitated by AI technologies.

The 90-Day Window Closed. Nobody Sent a Notice.
DISPATCH / MAY 2026 SECURITY · DISCLOSURE COLLAPSE · COMMIT MONITORING · PART 2
▲ Part 2 · Security Disclosure Closed · May 2026

Copy Fail’s mainline patch landed April 1. Public disclosure was April 29. The 28 days between commit and disclosure are the dangerous window — AI can rediscover the bug from the diff in minutes, while distribution patches take 2-8 weeks to reach end-user systems. Three asymmetries compound: time, expertise, knowledge category. Defender disadvantage compounds across all three.

▲ THE THREE ASYMMETRIES · ALL FAVOR THE ATTACKER NOW
Asymmetry 01
Time
90-day window collapses to diff-to-exploit minutes. Distribution lag becomes the structural vulnerability window.
Asymmetry 02
Expertise
5-10 year apprenticeship pipeline collapses to “find a security vulnerability” prompt + API access.
Asymmetry 03
Category
Memory safety → trust-boundary composition. Defensive infrastructure built for the wrong layer.
Defender disadvantage compounds across all three. Faster exploitation + more attackers + harder vulnerability category with less mature defense.
28days
Copy Fail · mainline commit → public disclosure
Apr 1 commit · Apr 29 disclosure · the dangerous window
$2M
Vercel customer data · BreachForums asking price
OAuth supply chain · Context.ai → Google Workspace
275M
Canvas records exfiltrated · ~9,000 institutions
ShinyHunters · Free-For-Teacher vulnerability · 3.65 TB
“find it”
Mythos prompt complexity · no security training
“Please find a security vulnerability in this program”
28-DAY WINDOW COPY FAIL MAINLINE COMMIT APR 1 → DISCLOSURE APR 29 · BUG REDISCOVERABLE FROM DIFF VERCEL APR 19 CONTEXT.AI → OAUTH → GOOGLE WORKSPACE → VERCEL ENV VARS → $2M BREACHFORUMS CANVAS MAY 1-12 SHINYHUNTERS · 275M RECORDS · 9,000 INSTITUTIONS · FINALS WEEK OUTAGE KNOWLEDGE FLOOR “PLEASE FIND A SECURITY VULNERABILITY” · NO TRAINING REQUIRED · ENGINEERS PRODUCED WORKING EXPLOITS DISTRIBUTION LAG MAINLINE → STABLE → DISTRO PACKAGE → DEPLOY · 2-8 WEEKS TYPICAL · LEGACY: NEVER CATEGORY SHIFT OAUTH SCOPES · SAAS TRUST · ENV VARS · FREE-TIER ABUSE · NOT MEMORY SAFETY 28-DAY WINDOW COPY FAIL · APR 1 COMMIT → APR 29 DISCLOSURE · BUG REDISCOVERABLE FROM DIFF
Asymmetry 01 · time · the commit-monitoring window

The patch is now the disclosure event.

Responsible disclosure orthodoxy: bug stays private until vendor patches. For open source, this has never been fully true — git commits are public in real-time. Copy Fail’s mainline patch landed April 1. Public disclosure was April 29. The 28 days between are the dangerous window.

Copy Fail · the disclosure-to-deployment timeline
Mainline commit is public from the moment it lands. Distribution propagation takes 2-8 weeks. AI processes the diff in minutes.
Apr 1 mainline ~Apr 10 stable Apr 29 disclosure Apr 30-May 7 distro patches +weeks deployed 28-day commit-to-disclosure window AI rediscovers from public diff PATCH IS PUBLIC · BUG IS PUBLIC · NO DEFENDER WARNING deployment lag unpatched systems exposed LONG TAIL · LEGACY · MONTHS+ AI watches every kernel commit “DOES THIS COMMIT FIX A SECURITY ISSUE?”
Apr 12026
Mainline commit lands. Linux kernel git tree publishes fafe0fa2995a reverting the 2017 in-place AEAD optimization. Patch is now public.
PUBLIC
INSTANT
~Apr 102026
Stable kernel backports. Greg KH’s stable trees include the patch. Still: no distribution package yet · no end-user deployment.
STABLE
TREES
Apr 292026
Public disclosure by Theori. CVE-2026-31431 announced. Most defenders learn of the bug 28 days after the patch was public on kernel.org.
CVE
PUBLIC
Apr 30 → May 72026
Distribution packages. Ubuntu, Amazon Linux, RHEL, SUSE, Debian, Fedora, Arch ship patched kernel packages. Each on its own schedule.
PACKAGES
AVAILABLE
+weeks → +months2026
End-user deployment. 30-day patch SLA · slower for regulated environments · effectively never for legacy systems without security updates.
DEPLOYED
SLOWLY
The 90-day window assumed private patches. Open-source patches are public from minute zero. The framework is misaligned with the capability landscape.
Asymmetry 02 · expertise · the knowledge floor collapse
Amazon

cybersecurity vulnerability scanner

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

“Please find a security vulnerability.”
No training required.

The historical pipeline for becoming a top-tier vulnerability researcher took 5-10 years of human apprenticeship. Kernel internals. Processor architecture. Exploit-mitigation-bypass craft. Decompiler-output reading. All baked into frontier model training data.

The knowledge floor · before AI / now
Who can do vulnerability research. Pool of capable actors expands by orders of magnitude.
▲ Before · 2015-2023
Senior researcher path
  • CS degree with security specialization
  • 3-5 years red team / CTF / firm experience
  • 2-3 years senior research with reportable findings
  • Tacit knowledge: kernel internals, decompiler output reading, exploit-mitigation-bypass craft
  • Global pool: ~200-500 senior researchers per decade
  • Apprenticeship: mentored by existing experts
▲ Now · 2026
API access + one prompt
  • Frontier model API access ($20-200/month for individuals)
  • One prompt: “Please find a security vulnerability”
  • No security training required (Anthropic / AISI / CETaS verified)
  • Tacit knowledge baked in from model training
  • Pool of capable actors: millions globally
  • Bottleneck: willingness to use it, not skill

The prompt Anthropic used to discover vulnerabilities with Mythos “essentially amounted to ‘Please find a security vulnerability in this program.'” Engineers with no formal security training were able to generate complete, working exploits.

— Alan Turing Institute · CETaS · Claude Mythos cybersecurity analysis
Asymmetry 03 · category · where the bugs actually live
Amazon

software patch management tools

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Memory safety isn’t where the breaches happen anymore.

Decades of defensive infrastructure built around memory safety (ASLR, NX bits, CFI, stack canaries). The most consequential breaches of April-May 2026 are not memory-safety bugs. They are trust-boundary failures at integration seams.

Two case studies · April-May 2026
No memory corruption. No kernel exploit. Trust-boundary composition failures. Mature defensive infrastructure for memory safety doesn’t apply here.

The bugs that matter most have shifted from memory safety to trust-boundary composition. OAuth scopes. SaaS-to-SaaS authentication. Multi-tier account models. Third-party app permissions. Environment variable handling. Defensive tooling for this layer is 5-7 years behind memory-safety discipline.

▲ CASE 01 · APR 19 2026
Vercel · the OAuth supply chain attack
$2MBreachForums asking price
Chain: Lumma Stealer infected Context.ai employee (Feb 2026) → harvested Google Workspace OAuth tokens → attacker used token to access Vercel employee Google Workspace → pivoted into Vercel account → enumerated and decrypted non-sensitive env variables → exfiltrated customer credentials → posted database on BreachForums.
Pattern: third-party AI tool → OAuth → identity → platform → customer secrets
▲ CASE 02 · APR 30 – MAY 12 2026
Canvas / Instructure · free-tier abuse + extortion
275Mrecords · 3.65 TB · ~9,000 institutions
Chain: ShinyHunters found vulnerability in Canvas Free-For-Teacher account mechanism → exfiltrated 3.65 TB across 275M records → ransom negotiations stalled → defaced ~330 institution login portals during finals week → school-by-school extortion through May 12. Names, emails, student IDs, private inbox messages exposed.
Pattern: free-tier authorization flaw → mass data exfiltration → multi-tier extortion

Defensive infrastructure for memory safety is 25+ years mature. Defensive infrastructure for trust-boundary composition is 5-7 years behind. AI-driven discovery operates at both layers — with less mature defenders at the layer that matters more for 2026 breaches.

Operational response · four audiences
Amazon

network security monitoring device

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

The defensive infrastructure that worked last decade doesn’t work at the same level now.

Adaptation is necessary. The 18-36 month window where defenders can build the necessary infrastructure is open. Asymmetric cost-of-being-wrong applies: capacity built is useful; capacity not built is structural vulnerability.

Operational response · by stakeholder
Calibrated to the new asymmetries · not to the historical defensive playbook.
▲ FOR CISOs
+ SECURITY TEAMS
Monitor upstream commits. Compress patch SLAs.
Implement upstream commit monitoring for kernels and critical software. Subscribe to mainline security lists. Evaluate suspicious commits with internal AI tooling. Target 72-hour deployment for kernel patches, 7-day for major apps, 14-day for everything else. Audit OAuth permission landscape. Treat SaaS supply chain as tier-1 infrastructure.
▲ FOR SOFTWARE
PUBLISHERS
Your commits document where your bugs are.
Security-shaped commits are findable by AI. Move toward private bug coordination for high-severity findings. Some vendors batch security fixes into general patches (Apple, Microsoft); open source structurally harder but worth attention. Run AI-driven discovery against your own codebase first — be first to know.
▲ FOR
POLICYMAKERS
Disclosure framework needs explicit policy attention.
Responsible disclosure is voluntary social technology that worked in the previous regime. Mandated disclosure standards, vendor patch SLA requirements, updated CVE management infrastructure. Linux distribution lag is a public-interest concern for critical infrastructure. OAuth/SaaS governance is a regulatory blind spot — Vercel is one of many March-April 2026 supply chain breaches.
▲ FOR
EVERYONE ELSE
Two-factor everything. Watch your OAuth grants.
Authenticator apps, not SMS. Passkeys where available. Aggressive credential rotation. Assume your SaaS providers will be breached — have a rotation playbook. Be wary of “Allow All” OAuth grants, especially for AI productivity tools requesting broad email/drive/calendar access. The Vercel chain started here.

The 90-day window collapsed. The knowledge floor collapsed. The bugs moved layers. Three asymmetries compound. The 18-36 month window where defenders can build the necessary infrastructure is open.

— Software security · the disclosure collapse · Part 2 · May 2026
Source dossier · the receipts
  • 732 Bytes to Root · the cost-curve collapse · Part 1
  • Theori / Xint Code · Copy Fail: 732 Bytes to Root · xint.io · Apr 29 2026
  • Linux kernel mainline patch · commit fafe0fa2995a · Apr 1 2026
  • CVE-2026-31431 · NVD · CVSS 7.8 (High) · CISA KEV listed
  • Project Zero · 90-day coordinated disclosure policy · 2014
  • Vercel Security Bulletin · April 2026 · vercel.com/kb/bulletin/vercel-april-2026-security-incident
  • Trend Micro · The Vercel Breach: OAuth Supply Chain Attack · Apr 21 2026
  • The Hacker News · Vercel Breach Tied to Context AI Hack
  • TechCrunch · Zack Whittaker · App host Vercel says it was hacked · Apr 20 2026
  • Hudson Rock · Context.ai Lumma Stealer compromise · Feb 2026
  • BleepingComputer · Vercel breach disclosure · Apr 19 2026
  • Instructure security incident · official disclosures · May 1-12 2026
  • Halcyon · Education Sector in the Crosshairs: ShinyHunters’ Extortion Campaign Against Instructure
  • Wikipedia · 2026 Canvas security incident · ongoing as of May 12 2026
  • CNN · Canvas hack: What we know · May 2026
  • Hackread · ShinyHunters Instructure + Vimeo breaches · May 2026
  • Anthropic Claude Mythos Preview System Card · Apr 7 2026
  • Alan Turing Institute / CETaS · Claude Mythos cybersecurity analysis
  • UK AI Security Institute · Mythos cyber capability evaluation
Colophon · Part 2

Set in Source Serif 4, IBM Plex Sans, & IBM Plex Mono. Security-advisory aesthetic. Free to embed with attribution.

thorstenmeyerai.com

Software security · the disclosure collapse · Part 2 of 2 · May 2026

28 days · 275M records · $2M · “find it”

Amazon

vulnerability disclosure software

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Implications of the Disclosed Window Closure

The expiration of the 90-day window without vendor notices indicates a potential shift where attackers may gain an advantage over defenders. The traditional model relied on the assumption that patches would be deployed promptly, and that the window between patch release and exploitation was manageable.

With AI systems capable of monitoring commits, reverse-engineering patches in minutes, and developing exploits rapidly, the period of vulnerability has effectively shortened to zero. This could lead to more widespread, faster exploitation of vulnerabilities, especially in critical infrastructure and cloud services.

Furthermore, the lack of vendor disclosures raises concerns about transparency and accountability, as organizations may remain unaware of vulnerabilities until they are exploited in the wild, increasing risk exposure across sectors.

Evolving Practices in Vulnerability Disclosure

The responsible disclosure framework was established over two decades ago to balance the interests of researchers and vendors, with the 90-day window serving as a negotiated compromise. It was predicated on assumptions that no longer hold in 2026.

Recent high-profile breaches, such as those involving Vercel and Canvas, demonstrate that many of the most critical vulnerabilities today are not memory-safety bugs but trust boundary failures at integration points like OAuth scopes and SaaS-to-SaaS communication. These vulnerabilities often remain undetected by traditional security measures and are exploited rapidly once discovered.

Meanwhile, AI-driven discovery tools, like those used by Theori and Anthropic, can surface zero-day vulnerabilities and develop exploits within minutes of a patch being released, rendering the traditional 90-day window obsolete.

“The collapse of the knowledge floor and the rapid development of exploits by AI tools mean the traditional 90-day window is no longer a defender’s advantage.”

— Thorsten Meyer

Unclear Impact on Future Security Practices

It remains uncertain whether vendors will begin to adopt more proactive disclosure or if new frameworks will emerge to address the accelerated pace of vulnerability discovery driven by AI. The long-term implications for cybersecurity policy and industry standards are still developing.

Next Steps for Vulnerability Management and Policy

Security industry leaders and policymakers are likely to reevaluate disclosure standards, possibly moving toward more transparent or real-time notification systems. Organizations should prepare for faster threat timelines and consider adopting AI-based monitoring tools to identify vulnerabilities proactively.

Further research and discussion are expected to clarify how to balance responsible disclosure with the realities of AI-enabled exploitation, shaping the future of cybersecurity protocols.

Key Questions

What does the end of the 90-day window mean for cybersecurity?

It suggests that vulnerabilities could be exploited more rapidly after discovery, reducing the time defenders have to respond and patch, especially as AI tools accelerate exploit development.

Why have vendors not issued notices for the recent Linux kernel patch?

It is not yet clear whether vendors are intentionally withholding notices, are unaware of the implications, or are reevaluating their disclosure strategies in light of AI-driven attack capabilities.

How does AI change the landscape of vulnerability discovery?

AI can analyze patches, reverse-engineer vulnerabilities, and develop exploits within minutes, collapsing traditional windows of opportunity for defenders.

Will the responsible disclosure model be replaced?

It is uncertain. Industry experts are debating whether new models or regulations are needed to address the accelerated pace of vulnerability discovery and exploitation.

What should organizations do now?

Organizations should enhance their monitoring capabilities, adopt AI-driven security tools, and prepare for faster response times to emerging vulnerabilities.

Source: ThorstenMeyerAI.com

You May Also Like

Behind the Scenes: What Is Google WM Max Llc?

Tackling the mystery of Google WM Max LLC begins with understanding its connection to Google's advertising services and calculated charges.

Budgeting Apps: How Tech Can Help You Manage Money

Simplify your finances with budgeting apps that offer insights and savings tips, but discover the hidden features that can transform your money management journey.

Document Scanners: Duplex, ADF, and OCR Explained Simply

A comprehensive guide to document scanners’ duplex, ADF, and OCR features reveals how they can revolutionize your workflow, but there’s more to discover.

The Channel Move: Anthropic, Wall Street, and the Acquisition of the Real Economy

Anthropic partners with major private equity firms in a $1.5 billion joint venture to embed AI directly into thousands of portfolio companies, transforming enterprise AI deployment.