// the investigation

The investigation

SEVDIS312 LLC's entry-level recruiting ad is the most explicitly permissive recruiting copy in the corpus — every barrier to entry has been removed by name. The ad: "Language skills not required. Driver's license not required. Only requirement is desire to work". The ad explicitly waives all three of the basic gatekeeping requirements a compliant subcontractor would enforce, leaving the worker's unilateral willingness to show up as the sole hiring criterion. No language, no DL, no documentation, no cert-screen. Desire. The ad is the minimum-viable hiring filter for an operator whose business model is absorbing workers the rest of the market has screened out.

The other SEVDIS312 ads round out the pattern. A recruiting post for the AT&T KAMO market (Kansas / Missouri) separates a "Top Hand" role (1 year experience + TTT1) from a "Green Guy" helper role ("no experience, or minimal experience") — the two-tier structure that allows experienced workers to be hired alongside complete entrants at calibrated pay differentials. A later SEVDIS312 post for Kansas / Missouri February start adds TTT1 and OSHA-30 as requirements for top hands, with pay two weeks after site close — a modestly more compliant variant of the same company's recruiting. The operator behind SEVDIS312 is Rafael, with phones 312-826-6491 and 786-833-5580 — both numbers shared with a separate Telegram account named "Leo," suggesting a single operator running two recruiting channels across the same phone infrastructure.

SEVDIS312 recruits for AT&T's KAMO build market. AT&T's vendor-management program could read this Telegram channel — it is the channel where SEVDIS312 publishes its no-language, no-DL, no-cert-screen entry-level offer. AT&T could check, against its own certification systems, whether the workers on SEVDIS312 closeouts are the workers actually hired through that ad and on what authorization. It has not. It knows.

// findings

3 findings on this card

Each finding is a single corpus message (or a short cluster) with a verbatim quote, severity tag, and provenance label. Click any finding to view the source message and context window. Speaker identifiers below are Telegram name (the handle the user chose for themselves on Telegram — can be a real name, a single letter, an emoji, or anything else they set) plus Telegram UID (the numeric account identifier).

m0259 HIGH JOB_POSTING_AMBIGUOUS PROVEN

Знание языка не обязательно Наличие DL не обязательно

Language skills not required Driver's license not required

source: CHAT · msg 46137 · Telegram name: ermkv_ · uid 6768314326
View source message and context →
m0258 MEDIUM JOB_POSTING_AMBIGUOUS PROVEN

Грин гай: без опыта,либо с минимальным опытом

Green guy: no experience, or minimal experience

source: CHAT · msgs 23809-37787 (3) · Telegram name: ermkv_ · uid 6768314326
View source message and context →
m0260 MEDIUM JOB_POSTING_LEGIT PROVEN

SEVDIS312 LLC набирает сотрудников

SEVDIS312 LLC is hiring

source: CHAT · msg 50332 · Telegram name: ermkv_ · uid 6768314326
View source message and context →
Source
Primary msg
Telegram Name
Telegram UID

CONTEXT WINDOW

Before you read

The Shadow Zone Investigation documents labor violations in U.S. wireless tower construction — including wage theft, undocumented-labor schemes, fraudulent safety certifications, child labor, and immigration coercion. 339 documented findings across 93 named subcontractors and 384 individuals. Anonymous investigation team.

Three editorial commitments:

  • Every finding carries a PROVEN, ALLEGED, or INFERRED label. The label is the editorial commitment, not decoration.
  • Russian-language source quotes are preserved verbatim, with English translations.
  • Three corruption allegations against named primes are framed as ALLEGED throughout. The corpus does not prove them. They are surfaced because named industry sources, separately, raised them. Every unproven allegation on this site is collected on the What Is Alleged — and Not Proven page.

Mistakes may exist. The corpus runs to ~30,000 messages over 5+ years; translations were AI-assisted across four independent analyses. Some errors likely remain. To report a factual error or dispute a finding's framing, see the corrections policy on the About page.