Oracle's OWBPA Leak: 546 Ages, a 52-Day Head Start on WARN
Oracle's leaked OWBPA disclosure lists 546 OCI layoffs by title and age. Here is how sourcers turn ADEA exhibits into named senior shortlists.
Business Insider's September 2026 scoop on Oracle's America Cloud Infrastructure cuts was reported as a layoff story. It is actually a sourcing document. The leaked exhibit lists 546 workers by job title and integer age, plus the retained population inside a 7,185-person decisional unit, because federal age-discrimination law required Oracle to hand every affected 40+ employee exactly that list.
If you source senior cloud engineers, this is the highest-signal artifact you will see all quarter. Here is what it contains, why it leaks, and how to work it before the WARN notice ever files.
What an OWBPA disclosure actually is
An OWBPA disclosure is a written exhibit an employer must give every laid-off worker aged 40 or older during a group termination, listing the job titles and integer ages of every person in the "decisional unit" who was selected for layoff and every person in that unit who was not. It exists to make ADEA (Age Discrimination in Employment Act) waivers enforceable, and a defective one voids the waiver.
The statute is 29 U.S.C. § 626(f)(1)(H). In practice, the document circulates under several names, which is what you set Google Alerts on:
- "Exhibit A"
- "ADEA attachment"
- "OWBPA Chart"
- "decisional unit list"
Two mechanics make this document leak reliably:
- Every affected employee legally holds a copy. In Oracle's case, that is 546 people, any one of whom can post it to TheLayoff, Blind, or hand it to a reporter.
- Plaintiffs' employment firms - Nisar Law, Sobel Law, Whitman Legal Solutions - publish "how to read your OWBPA disclosure" explainers to recruit claimants, because a defective disclosure resurrects an already-signed ADEA claim. That ecosystem incentivizes circulation.
The EEOC's severance-agreement Q&A page is the authoritative primary source and worth bookmarking before you argue with a compliance team about linking to one.
The Oracle OCI exhibit, by the numbers
Oracle's leaked September 2026 OWBPA disclosure covered 546 cuts inside a 7,185-person America Cloud Infrastructure org (7.6%), with Software Developer III as the single most-hit title at 57 roles. The demographic skew is what makes it a sourcing goldmine, not the count.
The paradox on the same page of Oracle's earnings deck: cloud infrastructure revenue hit $7.4 billion for the quarter, up 121% year over year, while the org running that revenue shipped a 7.6% RIF. A separate TIME / What We Will survey of 272 laid-off Oracle workers found 62% were 40 or older and 22% had 15+ years of tenure. This is not a junior cull. It is a tenure cull inside the fastest-growing revenue line the company reports.
Here is the dataset you should be working from:
| Slice | Count | Source / method |
|---|---|---|
| Oracle America Cloud Infrastructure decisional unit | 7,185 | Leaked OWBPA disclosure via Business Insider |
| OCI employees selected for layoff | 546 (7.6%) | Same disclosure |
| Software Developer III roles cut at Oracle OCI | 57 | Qz.com summary of BI document |
| US "Software Developer III" profiles, all employers | 1,100 | Refolk's index, exact-title match |
| US senior/principal engineers with "OCI" in headline | 203 | Refolk's index, seniority filter |
| US senior/principal engineers at Oracle mentioning OCI | 24 | Refolk's index, employer + keyword |
| Ratio: OCI cut cohort vs. visible senior OCI headcount | 2.7x | Derived (546 / 203) |
| Share of national Software Developer III pool cut in one exhibit | ~5.2% | Derived (57 / 1,100) |
Two numbers on that table should stop you. First, only 203 US senior or principal engineers list "OCI" in their headline anywhere in Refolk's index. Oracle's exhibit cut 2.7 times that visible population in one filing. Most of the affected engineers are effectively invisible to keyword search today, which is why LinkedIn Recruiter alone will not find them. Second, 57 Software Developer IIIs at one company is about 5% of the entire named national pool at that exact title.
Why OWBPA beats WARN for senior IC sourcing
WARN notices tell you location and count. OWBPA disclosures tell you title, seniority, and integer age, which is what you actually need to pick people. That is the whole difference.
WARN was designed for plant closures. It is optimized for advance notice to state workforce agencies about how many people at what address, and it lags. Medhacloud's tracking noted more than 3,000 Oracle roles eliminated across the US, India, Philippines, and Canada by mid-2025 that showed up in state WARN filings without any accompanying company announcement. Weeks late, no titles.
An OWBPA exhibit gives you three things WARN never does:
- Exact job titles. "Software Developer III" versus "Software Developer IV" is the difference between two very different comp bands and tenure ranges.
- Integer ages. Age 47 with a Software Developer IV title cross-references to a joined-Oracle date range that narrows LinkedIn candidates per row to single digits.
- The retained list. More on this below. WARN never tells you who is still there.
For sourcing senior engineers from layoffs, this changes the ranking of your inputs. OWBPA disclosure beats company press release beats state WARN filing beats Blind rumor thread. Most sourcing teams still run them in the opposite order.
The move nobody makes: source the survivors
The retained side of an OWBPA disclosure is the most valuable half, and almost no recruiter works it. A group-layoff exhibit lists ages and titles of everyone selected AND everyone in the same decisional unit not selected, which means a leaked Oracle document names the 6,639 OCI engineers who kept their jobs alongside the 546 who did not.
Those 6,639 people are doing more work at worse morale, watching peers get walked out, and receiving a very clear signal about who management considers expendable at their title band. That is a warmer passive-candidate segment than the freshly-laid-off, who are fielding fifty inbound messages and probably taking severance to breathe for a month.
The exhibit tells you, per title band, exactly how many senior ICs Oracle kept and what their age distribution looks like. Combine that with a title-plus-tenure inference against a real profile index and you get a named survivor list. This is the exact gap Refolk closes: you describe the person in plain English ("senior distributed systems engineers still at Oracle OCI, joined 2015 to 2019, based in the US") and get a ranked shortlist, without hand-mapping 6,639 rows to LinkedIn URLs.
The retained side of an OWBPA exhibit is a warmer list than the cut side. Almost nobody works it.
The 45-day sourcing calendar
An OWBPA disclosure is not a point-in-time event. It is a 52-day sequencing calendar, and treating it as a single date is why most sourcers waste it.
Federal law requires:
- A 45-day consideration window for group layoffs before an affected employee can sign the ADEA waiver.
- A 7-day revocation window after signing, during which the agreement is not final.
That gives you a 52-day arc from disclosure to legal finality per employee. Practically:
- Days 0 to 10. Employee is stunned, reading the exhibit, calling a lawyer. Do not cold-outreach here; it converts terribly and it looks predatory.
- Days 10 to 35. Employee is still legally employed, demoralized, and starting to run numbers on severance versus a new offer. This is your best outreach window. A warm intro from an ex-colleague on the retained list beats a cold recruiter InMail.
- Days 35 to 52. Signing and revocation. Employee is now firmly deciding between severance-plus-search and a bird-in-hand offer. Comp negotiations shift.
- Day 52+. The market catches up. WARN filings post, LinkedIn "Open to Work" banners appear, inbound spikes, your leverage evaporates.
The named-shortlist advantage from a leaked exhibit lasts roughly seven weeks. That is your window.
Turning an exhibit into named candidates
Title plus integer age gives you tenure inference; tenure inference plus employer plus location narrows to single-digit LinkedIn candidates per exhibit row. Then you verify with GitHub or conference talk history.
A worked example on one Oracle exhibit row, illustrating the method:
- Exhibit row: "Software Developer IV, age 43, America Cloud Infrastructure."
- Tenure inference: Oracle promotes to Developer IV at roughly 6 to 9 years of company tenure. Joined-Oracle window: 2017 to 2020.
- Employer plus keyword filter across a real index: current title Software Developer IV, employer Oracle, headline mentions OCI, US-based.
- Age filter: age 43 brackets graduation year 2003 to 2006, which narrows the candidate set again.
- Result on this synthetic row: typically 2 to 6 named profiles, verifiable against GitHub commit history to Oracle-owned repos or conference speaker bios.
Do that 546 times and you have a named shortlist, not a demographic summary. Do it against the retained side and you have 6,639 named survivors, ranked by how squeezed they probably feel.
The hardest part is not the exhibit. It is the fan-out to profiles. When only 24 US senior or principal engineers at Oracle mention OCI in their headline anywhere in Refolk's index, you learn that headline keyword filtering alone loses you roughly 90% of the real population. You need an index that reads work history, project descriptions, GitHub org membership, and conference talks, not just current-title strings. That is where "ask for the people you want and get them back" beats hand-crafted boolean strings: the semantic layer catches the 179 senior OCI engineers who never bothered to put "OCI" in their headline.
Where the exhibits actually leak
Watch four surfaces, in this order of signal:
- Plaintiffs'-side employment law blogs. Nisar Law, Sobel Law, Whitman Legal Solutions, and the layoffcalculator.com insights hub publish anonymized OWBPA exhibits to attract claimants. Set alerts on the four canonical exhibit names above plus the employer.
- TheLayoff.com and Blind. Verified employees post exhibits during the 45-day window, often with rich thread commentary on which teams were hit. The Oracle threads were last updated September 14, 2026, mid-window.
- Business Insider and similar outlets. Recurring recipients of tech-layoff exhibits, publishing enough detail (org name, count, median age) to trigger a targeted sourcing pass even without the raw doc.
- State WARN filings. Useful for confirmation and cross-referencing location, but always trailing the disclosure by weeks.
If a company you recruit against announces a "restructuring" affecting more than one 40+ employee, an OWBPA disclosure exists. It has been printed and handed to at least two people. Your job is to find it before day 35.
FAQ
Is it legal to source from a leaked OWBPA disclosure?
Yes, with normal caveats. The document is a private communication between employer and employees, but once it circulates on TheLayoff, in press coverage, or on a plaintiffs' law firm blog, using its public contents to identify candidates is not different from using any other public disclosure. The tortious-interference risk is around contacting people who are still employed under a non-solicit; the 45-day consideration window is exactly when they legally still are. Talk to your employment counsel about your specific outreach copy, particularly if the target signed a non-solicit as part of the severance package.
How is an OWBPA disclosure different from a WARN notice?
WARN notices are filed with state workforce agencies and disclose employer name, affected site address, total headcount reduction, and effective date. OWBPA disclosures are handed to individual employees and disclose job titles and integer ages for the selected and retained populations inside the decisional unit. WARN tells you where and how many. OWBPA tells you who, at what seniority, and how old.
What is the "decisional unit" and why does it matter for sourcing?
The decisional unit is the population from which the employer chose who to lay off, typically an org, a business unit, or a geographic site. It matters because the exhibit lists everyone in that unit, cut and kept, which lets you reconstruct the org's full title-and-age distribution. For Oracle's September 2026 exhibit, the decisional unit was the 7,185-person America Cloud Infrastructure org, giving you a demographic map of the entire OCI Americas engineering population, not just the 546 people who left.
Why is the retained list more valuable than the cut list?
Cut candidates are fielding heavy inbound within days and often take severance to decompress before their next move. Retained candidates are still employed, doing more work, watching their peers get walked out, and receiving a clear signal about how management values their title band. Warm passive outreach to the retained population converts better than cold outreach to the freshly-laid-off, and the negotiation dynamics favor the recruiter because the candidate is not choosing between severance and your offer. Almost nobody works this side of the exhibit, which is why it does.
Try it on the search you came here for
Stop building boolean strings. Just describe the person.
Type one sentence. I plan the search, read GitHub, public LinkedIn and Crunchbase records, and the open web as it is right now, and hand back a ranked list with the reason next to every name.
01Describe them
One plain sentence. Role, city, stack, stage, whatever matters to you.
02I read the web live
GitHub, public LinkedIn and Crunchbase records, the open web. Not a database that went stale last quarter.
03You read the shortlist
Ranked, with the reasoning under every name. Open a profile, ask a follow-up, narrow it down.
- Staff backend engineers in NYC who shipped Rust in production
- Series A fintechs in SF under 50 people, growing headcount this year
- Maintainers of fast-growing Rust web frameworks on GitHub
- No boolean, no filters, no seat to buy. One box.
- Read at search time, so a profile updated yesterday counts today.
- Every step visible as it runs, every name with its reason.
500 free credits on sign-up. No card, no demo call. See real searches.