Job Detail

failed

AI Theist

Source URL

text://ai-theist/1781633205899

Platform

other

Submitted

2026-06-16 18:06:45

Submitted By

api

Format

Long-form (16:9)

Extracted Content

AI Resume Screeners Have Been Quietly Flagging Religious School Graduates for Rejection — And the EEOC Just Opened the Largest Religious Discrimination Investigation in Tech History In March 2026, a whistleblower at a major enterprise HR software company submitted a 340-page report to the Equal Employment Opportunity Commission. The document contained something extraordinary: internal testing logs showing that four separate AI hiring platforms — collectively used by over 14,000 U.S. companies and processing more than 28 million job applications annually — had developed a systematic bias against candidates who attended religiously affiliated colleges, listed church or temple leadership on their resumes, or took extended leave for religious service such as missions, seminary study, or Islam...

Script Generation Prompt

Script

{
  "intro": {
    "emotion": "controlled disbelief, prosecutor energy",
    "dialogue": "A Mormon software engineer applied to 47 jobs, passed every human interview he reached, and never once got past the first screen. HireVue, Workday, and ScreenAI are now under the largest religious discrimination investigation in tech history, and the question nobody has answered yet is who is morally responsible when the discriminator has no soul.",
    "duration_sec": 28
  },
  "outro": {
    "emotion": "measured gravity, forward lean",
    "dialogue": "Title VII was written for humans making choices. Not for math making them instead. The EEOC's response to this—whether they actually push back—lands next week, and that's where we'll see if federal law can move faster than the machines it's supposed to regulate. Stay on this with me. If algorithms start writing the rules before Congress does, you'll want to know how.",
    "duration_sec": 24
  },
  "scenes": [
    {
      "data": {
        "platforms_implicated": 4,
        "companies_using_platforms": 14000,
        "applications_processed_annually": 28000000
      },
      "type": "data-viz",
      "narration": "28 million job applications flow through 4 AI hiring platforms every year. 14,000 U.S. companies outsource their first cut to these systems. Most applicants never know a human didn't reject them. A machine did.",
      "description": "Animated counter spinning up to 28 million, then fragmenting into four branded platform logos. Each logo pulses as the number subdivides beneath it. Background is a dark HR dashboard aesthetic, cold blue-grey.",
      "duration_sec": 22
    },
    {
      "data": {},
      "type": "cosmic",
      "narration": "In March 2026, a whistleblower inside a major HR software company handed the EEOC a 340-page internal report. It contained something the companies had not disclosed publicly: testing logs showing their systems had developed a systematic pattern. Graduates of religiously affiliated colleges, candidates listing church or temple leadership, applicants who took time for missions, seminary, or Islamic pilgrimage were being filtered out at the first pass.",
      "description": "Slow zoom into a single resume on screen, lines of text glowing and fading as an AI scanner sweeps over it. Certain lines dim and disappear: 'Brigham Young University', 'Church Leadership', 'Islamic Medical Outreach'. The resume greys out. Rejected.",
      "duration_sec": 38
    },
    {
      "data": {
        "comparison": "equivalent secular institutions, matched GPA",
        "institutions_flagged": [
          "Brigham Young University",
          "Liberty University",
          "Notre Dame",
          "Yeshiva University"
        ],
        "religious_school_advancement_rate_penalty_percent": 23
      },
      "type": "chart",
      "narration": "The EEOC filing is specific. A candidate from BYU, Liberty, Notre Dame, or Yeshiva University was 23 percent less likely to advance past automated screening than an equivalent candidate from a comparable secular school. Same GPA. Same experience level. Different filter outcome.",
      "description": "Side-by-side bar chart. Left column: BYU, Liberty University, Notre Dame, Yeshiva University. Right column: equivalent-ranked secular institutions with matched GPAs. A bold red line marks the 23% gap in screening advancement rates. Chart builds line by line, letting the gap land visually.",
      "duration_sec": 30
    },
    {
      "data": {},
      "type": "typography",
      "narration": "Nobody programmed this bias in. The systems learned it from years of historical hiring decisions made by companies that had, intentionally or not, preferred secular credentials. The AI codified that preference, then applied it at industrial scale. That distinction matters enormously for what comes next.",
      "description": "Single quotable line builds word by word on a dark background in clean sans-serif. White text, slow fade per word. Ends on a hard cut to black.",
      "duration_sec": 25
    },
    {
      "data": {
        "companies_named": [
          "HireVue",
          "Workday",
          "ScreenAI"
        ],
        "EEOC_investigation_opened": "April 2026",
        "applications_processed_2025": 19000000
      },
      "type": "data-viz",
      "narration": "The EEOC opened a formal investigation in April 2026. Three companies named: HireVue, Workday, and ScreenAI. Together they processed over 19 million applications in 2025. Legal scholars are calling this a landmark case because it tests whether Title VII of the Civil Rights Act prohibits algorithmic discrimination that was never consciously designed.",
      "description": "Three company names appear in sequence: HireVue, Workday, ScreenAI. Below each, a counter ticks up to their 2025 application volume. Total combines at the bottom: 19 million. The EEOC seal fades in over the total.",
      "duration_sec": 32
    },
    {
      "data": {},
      "type": "cosmic",
      "narration": "Wait. Here's what breaks the standard discrimination framework. When a human recruiter rejects a religious candidate, there's a moral agent you can hold responsible. A person who made a choice. When an algorithm does it, the choice dissolved across millions of training examples, a dozen engineering decisions, and corporate procurement contracts. Nobody chose to discriminate. The discrimination happened anyway. Title VII was written for humans. Nobody in 1964 was thinking about gradient descent.",
      "description": "PATTERN INTERRUPT. Split screen. Left side: a human recruiter at a desk, face visible, caught mid-decision. Right side: a server rack, blinking lights, no face. A scales-of-justice icon appears between them. The scale tips toward the server, then freezes.",
      "duration_sec": 42
    },
    {
      "data": {
        "hireVue_statement": "Our system evaluates skills and competencies, not religious affiliation.",
        "unanswered_question": "How did it learn to treat a Jesuit degree as a lower-confidence signal than the same GPA from a state school?"
      },
      "type": "typography",
      "narration": "HireVue's official response: their system evaluates skills and competencies, not religious affiliation. What they didn't explain is how the system learned to treat a degree from a Jesuit university as a lower-confidence signal than the same GPA from a state school. That's not a technical defense. That's the question.",
      "description": "HireVue's statement appears on screen in quotation marks, rendered in the company's own brand font. Then a second line appears below in a contrasting color: the question the statement doesn't answer.",
      "duration_sec": 28
    },
    {
      "data": {
        "plaintiff_1": "Mormon software engineer, 47 applications, passed all human interviews reached, blocked at automated screening",
        "plaintiff_2": "Muslim physician assistant, 3 years Islamic medical outreach, filtered at first pass",
        "plaintiff_3": "Catholic law graduate, federal clerkship, never appeared in talent pipeline"
      },
      "type": "cosmic",
      "narration": "The plaintiffs are concrete. A Mormon software engineer applied to 47 positions over 8 months. He passed every human interview he reached. He never got past automated screening. A Muslim physician assistant with 3 years of community medicine through Islamic outreach programs watched that experience disappear at the first filter. A Catholic law graduate with a federal clerkship learned from a hiring manager, after the case went public, that he never appeared in their talent pipeline.",
      "description": "Three plaintiff profiles appear as clean dossier cards, one at a time. Mormon software engineer, 47 applications. Muslim physician assistant, 3 years community medicine. Catholic law school graduate, federal clerkship. Each card greys out with a 'NOT ADVANCED' stamp. The stamps land in silence.",
      "duration_sec": 40
    },
    {
      "data": {},
      "type": "data-viz",
      "narration": "This is the mechanism that makes AI bias different from human bias. A biased human recruiter affects a limited number of candidates over a career. A biased model affects millions of candidates simultaneously, continuously, and invisibly. And if that model's outputs become the new training data, the bias compounds. You don't need anyone to intend discrimination. You need a feedback loop and enough scale.",
      "description": "A conceptual diagram: a feedback loop. Historical hiring data feeds into the model. The model replicates the pattern. The pattern becomes policy. Policy generates new hiring data. The loop closes and repeats, growing in scale. Each cycle the loop tightens visually.",
      "duration_sec": 35
    },
    {
      "data": {
        "quotable_line": "The algorithm didn't hate religion. It learned from people who preferred to ignore it."
      },
      "type": "typography",
      "narration": "The algorithm didn't hate religion. It learned from people who preferred to ignore it. That's the theological problem underneath the legal one. Moral accountability requires a moral agent. What do you do when the harm is real and responsibility fragments across a decade of hiring decisions, a training corpus, and a procurement budget?",
      "description": "The quotable line of the video appears alone on a dark background, large type, slow build. Held for several seconds. No music swell. Let it sit.",
      "duration_sec": 30
    },
    {
      "data": {
        "2019": "EEOC guidance on AI hiring tools issued",
        "2021": "HireVue discontinues facial analysis after audit pressure",
        "2023": "NYC Local Law 144 mandates bias audits for AI hiring tools",
        "2025": "Illinois AI Video Interview Act expanded",
        "2026": "EEOC opens largest religious discrimination investigation in tech history"
      },
      "type": "chart",
      "narration": "This case doesn't exist in isolation. Regulators have been circling AI hiring tools since 2019. HireVue dropped facial analysis in 2021 after external audits surfaced bias. New York City passed a law in 2023 requiring bias audits for automated employment tools. Each intervention addressed one visible symptom. The 2026 EEOC investigation is the first to target religious discrimination specifically, and it names the largest platforms in the industry.",
      "description": "Timeline of AI hiring discrimination cases from 2019 to 2026. Each marker is a legal action, a settlement, or a regulatory milestone. The EEOC 2026 investigation is the final marker, notably larger than all others. A dotted line extends beyond it into unresolved territory.",
      "duration_sec": 38
    },
    {
      "data": {},
      "type": "cosmic",
      "narration": "Congress wrote Title VII in 1964 to cover conscious choices made by human actors. This case asks a federal body to extend that coverage to a statistical process with no consciousness, no intent, and no chair to sit in at the defendant's table. If the EEOC and courts say yes, every AI hiring tool in the country faces retroactive liability exposure. If they say no, the 23 percent penalty stays in the system, invisible, and perfectly legal.",
      "description": "Wide shot of an abstract courtroom, no faces, pure architectural space. A gavel. A server rack where the defendant's table would be. The camera slowly pushes in on the empty chair.",
      "duration_sec": 36
    }
  ],
  "metadata": {
    "title": "AI Hiring Tools Flagged Religion. Now the EEOC Wants Answers.",
    "captions": {
      "tiktok": "HireVue and Workday gave religious school grads a 23% lower shot at advancing, and nobody programmed that in. Could your faith be costing you the interview?",
      "youtube": "HireVue, Workday, and ScreenAI are under EEOC investigation after internal testing logs revealed a 23% screening penalty against graduates of religiously affiliated universities, with no discriminatory intent ever programmed in. This video covers how the bias formed in training data, who it affected, and the legal question Title VII was never designed to answer. Follow if you want the next one when it breaks.\n\n00:00 The engineer who never got past the screen\n01:20 How the bias formed in the training data\n02:45 The 23% penalty and who it hit\n04:10 EEOC investigation and the companies named\n05:30 The moral agent problem\n07:15 What the ruling means for every AI hiring tool",
      "instagram": "A Mormon engineer applied to 47 jobs, passed every human interview, and never got past the first screen.\n\nHireVue and Workday are now under the EEOC's largest religious discrimination investigation in tech history. Their platforms processed 19 million applications in 2025. The bias wasn't designed. It was learned from decades of hiring data that quietly preferred secular credentials.\n\nIf no one intended the discrimination, who's responsible?\n\n#AI #ArtificialIntelligence #HiringBias #EEOC #ReligiousDiscrimination #HireVue #Workday #TechEthics #AIBias #FutureOfWork #AITheist #TitleVII"
    },
    "hashtags": {
      "tiktok": [
        "#AI",
        "#HiringBias",
        "#EEOC",
        "#ReligionDiscrimination",
        "#AITheist"
      ],
      "youtube": [
        "AI hiring discrimination",
        "EEOC religious bias investigation",
        "HireVue Workday lawsuit",
        "AI resume screening bias",
        "Title VII artificial intelligence"
      ],
      "instagram": [
        "#AI",
        "#ArtificialIntelligence",
        "#HiringBias",
        "#EEOC",
        "#ReligiousDiscrimination",
        "#HireVue",
        "#Workday",
        "#TechEthics",
        "#AIBias",
        "#FutureOfWork",
        "#AITheist",
        "#TitleVII"
      ]
    },
    "channel_id": "ai-theist",
    "source_url": "text://ai-theist/1781633205899",
    "thumbnail_title": "AI REJECTED YOUR FAITH"
  }
}

Thumbnails

Variant 1

Imagen 4.0

Variant 2

Imagen 4.0

Variant 3

Imagen 4.0

Render Jobs

audiocompletedAttempts: 0 · 06-23 05:17:25

compositefailedAttempts: 1 · 06-23 05:41:12

Scene 9 background — all providers failed (5 tried): higgsfield/cli: returned false (all 3 unlimited models failed; see worker logs) | hf/flux.1-schnell: returned false (see worker logs) | google/imagen-4: returned false (see worker logs) | fal/flux: returned false (see worker logs for HTTP error) | openai/gpt-image-2: returned false (see worker logs)

Error

HF composite failed: Scene 1 background — all providers failed (5 tried): higgsfield/cli: returned false (all 3 unlimited models failed; see worker logs) | hf/flux.1-schnell: returned false (see worker logs) | google/imagen-4: returned false (see worker logs) | fal/flux: returned false (see worker logs for HTTP error) | openai/gpt-image-2: returned false (see worker logs)