If you would rather read than watch the full interview above, this Reader’s Cut below carries the complete argument, evidence, and factual corrections in text.
The sentence that stayed with me after my conversation with Ashley E. Jones was not about a chatbot, a billionaire, or some promised artificial superintelligence. It was her description of Black women as the country’s best risk managers. Black women, she argued, are often the first to recognize the harm coming through a workplace, a political movement, or a new technology because we are usually among the first people expected to absorb it.
That observation tied together a conversation that moved from Dr. Timnit Gebru’s warnings about large language models to Laura Loomer’s public attacks on Black women, from Jason Black’s contempt for Black women’s freedom to a lawsuit challenging Workday’s hiring tools. These subjects can look unrelated when the internet chops them into clips. They are connected by a question that institutions keep avoiding: What happens when the person who identifies the risk has less power than the person creating it?
Gloria Steinem’s death at 92 gave the broadcasting an unexpected historical frame. Her place in the women’s movement matters, but neither Jones nor I wanted one famous white woman to become the whole history of feminism. The more urgent issue was what solidarity requires now, especially when Black women are expected to warn everybody, protect everybody, and then accept being treated as the problem.1
TLDR
Bias does not begin inside the machine. Large language models learn from human-created data, which means the hierarchies, exclusions, and stereotypes of the source material can travel into the system at scale.2
Black women are treated as alarms that can be fired. Jones connected Gebru’s exit from Google to a familiar workplace pattern: an institution hires a Black woman to identify risk, then resents her when the warning reaches the institution’s power center.3
Misogynoir supplies cultural training data. The rhetoric aimed at Black women by Laura Loomer and Jason Black does more than insult individuals. It reduces Black women to racial and sexual types that can be dismissed before their arguments are heard.1
Automated decisions do not erase responsibility. The Workday litigation remains unresolved, but courts have allowed significant discrimination claims to proceed, and the EEOC has argued that an AI vendor may be covered by federal anti-discrimination law when its tools effectively act for employers.4
An audit trail is not the same as accountability. Flock says searches in its license-plate-reader system are logged, yet documented cases show that officers have been accused of using surveillance databases to monitor women for personal reasons.5
Governance has to arrive before deployment. Risk assessment, independent audits, narrow access, enforceable consequences, and meaningful authority for affected communities cannot be optional repairs after harm occurs.
Restack this report, share it, and send it to one person who still believes technology can somehow rise above the politics, labor, and prejudices of the people who build it. The evidence is easier to ignore when it remains trapped inside a livestream.
Ashley E. Jones on Misogynoir, AI and Power: Watch by Chapter
Tap any timestamp to jump directly to that moment in the full broadcast.
1 | THE FRAME AND THE HISTORY
0:00 — Timnit Gebru on the limits of dominant AI research
0:36 — XVOA opens the broadcast
0:46 — Ashley E. Jones joins the conversation
2:20 — Sovereign Made Co. and governance-forward technology
3:57 — Remembering Gloria Steinem without erasing other women
5:39 — Black women inside the history of feminism
2 | WHO BUILDS THE MACHINE
7:09 — Timnit Gebru on hegemonic data
7:44 — How human bias enters AI training data
11:44 — The social cost of one corporate AI future
12:34 — Retaliation against Black women who identify risk
16:20 — How institutions devalue Black women’s warnings
19:21 — Timnit Gebru: technology is political
19:59 — Why technology cannot escape politics
22:23 — Tech leadership without independent accountability
25:46 — The audience enters the discussion
3 | MISOGYNOIR AS TRAINING DATA
26:24 — Laura Loomer attacks Black women in public life
27:14 — Misogynoir and dehumanizing racial tropes
30:08 — Loomer’s written attack
30:22 — Racist rhetoric becomes a political weapon
31:17 — Jason Black attacks Black women commentators
33:31 — Contempt presented as political analysis
34:45 — Black women described as post-emancipation “free radicals”
35:09 — When Black women’s freedom becomes the alleged threat
37:17 — Womanism and the limits of mainstream feminism
40:08 — Solidarity, discomfort and the live audience
4 | WHEN SOFTWARE MAKES DECISIONS
45:22 — Ashley Jones’s “Two Can Play That Game”
47:43 — The Workday lawsuit and algorithmic hiring bias
53:51 — AI, employment data and pay discrimination
55:36 — Flock cameras and surveillance abuse
1:01:15 — What accountability for surveillance should require
1:04:36 — Misusing law-enforcement databases
1:07:21 — Facial-recognition failures and Black faces
1:08:10 — Competing with China without abandoning Americans
1:10:18 — Technology companies, privacy and child safety
5 | WRITING, EDUCATION AND THE LINE
1:12:23 — AI allegations against Black authors
1:16:15 — Should young students use generative AI?
1:20:08 — Final questions and closing thoughts
The Warning Becomes the Offense
Ashley E Jones is a Chicago based strategist and writer who works at the intersection of artificial intelligence, human resources, and business operations. She founded Sovereign Made Co., which she describes as an intelligent-infrastructure company built around governance. That word, governance, is important because the public conversation about AI is still dominated by capability: what the model can write, imitate, classify, predict, or automate. Jones keeps returning to a different question: who authorized it to act, and who has the power to stop it?
Her answer begins with the people institutions usually place closest to harm but farthest from final authority. She called Black women “the best risk managers in the country,” not because of some mystical capacity, but because survival requires pattern recognition. The employee who notices the discriminatory shortcut, the patient who recognizes that a protocol was not built with her body in mind, or the citizen who knows a new surveillance power will not be used evenly is performing risk analysis before the institution has admitted that a risk exists.
The trouble begins when the warning touches revenue, prestige, or executive control. A company may celebrate diversity while the conversation stays ceremonial. Once a Black woman says the product itself is unsafe, the data is distorted, or leadership has misunderstood the people who will bear the consequences, her expertise can be recoded as hostility. The institution hires the alarm and then punishes it for making noise.
The Internet Is Not All Human Knowledge
The broadcast opened with Gebru challenging the assumption that a massive dataset represents humanity merely because it is large. The internet is not a neutral archive of every culture, language, and experience. Access is unequal, participation is unequal, and the most visible material often reflects the priorities of people with the greatest technical and economic power. The 2021 “Stochastic Parrots” paper that Gebru coauthored warned about the environmental costs, undocumented training data, embedded bias, and false appearance of meaning produced by very large language models.2
Jones translated that research into a plain institutional problem. If similar people build the systems, choose the data, define success, and judge the results, their worldview will not remain outside the model. It becomes part of the model’s operating environment. Diversity at the bottom of an organization does not solve this when dissent cannot alter the release schedule, stop a deployment, or force an audit.
Gebru’s departure from Google remains a disputed corporate episode in its particulars. Gebru said she was fired after the company demanded that a critical paper be withdrawn or stripped of the names of Google employees; Google described the separation differently. What is not disputed is that she left in 2020 after a conflict over research that challenged the direction of large-scale language technology.3 That history gave Jones’s risk-manager argument its sharpest example: expertise is welcome until it threatens the institution that employs it.
Misogynoir Does Not Stay in the Comment Section
Dr. Moya Bailey coined “misogynoir” to name the specific combination of anti-Black racism and misogyny directed at Black women. The broadcast examined two different examples. One came from far-right activist Laura Loomer, who used degrading racial language to attack Black women in public life.
The other came from Jason Black, host of The Black Authority, a long-form Black political-commentary channel launched in 2009 that had approximately 181,000 subscribers at publication. He is an established voice within a segment of Black YouTube where politics, economics, Black nationalism, and male-female “dynamics” are discussed in broadcasts that can run for hours. That reach is why his rhetoric belongs in the record. He is not some anonymous account shouting into an empty comment section; he has an audience, an archive, and a media platform capable of presenting contempt for Black women as serious political instruction.
The most revealing Jason Black clip was not merely an insult. It was a historical argument in which enslaved Black women were described as property who had once been controlled and contained, while emancipation turned them into “free radicals.” That language identifies the alleged disorder as Black female freedom itself. Once that premise is accepted, a Black woman who speaks, organizes, refuses male authority, or possesses a public platform can be treated as evidence that something has gone wrong.
Jones’ response was both political and personal. Loomer’s targets, she said, were her sisters because the attack was aimed at a category larger than any one official. Her analysis also exposed why political misogynoir belongs in a discussion about AI. Systems do not need to “hate” Black women in a human sense to reproduce a culture’s hierarchy. They need only be trained, tuned, evaluated, or deployed inside institutions where the hierarchy has already been normalized.
The Workday Case and the Disappearing Decision-Maker
Jones’ essay “Two Can Play That Game” examines how applicants can use AI to analyze job descriptions, identify gaps in a résumé, and better understand the screening systems standing between them and a human reviewer. The strategy is practical, but it carries an ugly irony. People who suspect that an automated system is screening them unfairly may need another automated system simply to make themselves legible to it.
Jones brought up the Workday lawsuit because it sits at the center of what she has been writing about which is that job applicants using AI to navigate automated systems may be getting screening out. Her essay “Two Can Play That Game” describes using AI to analyze a job description against a résumé, expose gaps, and translate legitimate experience into language an applicant-tracking system is more likely to recognize. She calls the strategy “using AI to fight AI.” The immediate objective is painfully modest: survive the automated screening layer long enough for an actual person to see the application.
The lawsuit Jones referenced was filed by Derek Mobley, who alleged that Workday’s algorithmic tools discriminated against applicants based on race, age, and disability. The EEOC has not declared those allegations proven. Its amicus brief argued that a software vendor may qualify as an employment agency or an agent of employers when its tools perform consequential screening functions. Courts have allowed important portions of the case to proceed, while Workday denies that its products make discriminatory employment decisions.
Jones then pushed the problem beyond the algorithm. “Who is the human reviewing the résumé?” she asked. Getting past the software does not guarantee fair consideration if the recruiter inherits the same assumptions, trusts the machine’s ranking, or believes an automated rejection must be objective. Human review can become another layer of the same architecture rather than a safeguard against it.
Her larger point was that AI did not descend upon a clean employment system. “AI is not new,” she said. “They just put it on top of existing systems and architecture that was built by humans.” An employer selected the system, a vendor designed it, managers established the process, and recruiters acted on its results. Jones was not describing a decision with no owner. She was describing a chain of owners who can each point somewhere else when the outcome is challenged.
The Database Knows Who Asked
I raised Flock cameras with Jones because my years in law enforcement taught me what access to a criminal-justice database can make possible. Training expired, credentials could lapse, and an officer was supposed to be able to connect a search to legitimate official work. Flock’s network extends that power by allowing participating agencies to search recorded vehicle sightings. Jones immediately moved past what the technology could find and asked the question that governs every surveillance system: “Who is the one accessing that capability?”
Her concern was not abstract. She asked who built the software, what safeguards had been designed specifically for women, who could examine error rates, and whether anyone outside the deploying agency had meaningful auditing authority. “Just because you’re a police officer doesn’t mean you get to stalk your ex-wife,” she said. Flock says its searches are tied to individual users, preserved in audit logs, and available for agency review. A searchable log can preserve what happened, but accountability still depends upon somebody examining it before private curiosity becomes official surveillance.
A 2026 Washington Post analysis gave her concern a documented foundation. The newspaper found at least 50 law-enforcement officers who had been charged with or accused of misusing license-plate-reader and other databases, including to monitor women without their knowledge or consent. That number covered multiple systems and should not be misrepresented as 50 proven Flock violations. It nevertheless shows why Jones kept returning to access and enforcement. A department may possess an audit trail while leaving detection to a supervisor, a victim’s complaint, or the scandal that arrives after the surveillance has already occurred.
Jones called attention to the person with “god-level access,” the administrator or connected insider who can reach information ordinary people cannot. For her, the question is no longer simply how somebody obtained private information. It is who granted the access, what qualifications that person possessed, and what consequence followed when the privilege was abused. When public safety depends upon privately built infrastructure, government remains responsible for controlling another powerful actor.
She applied the same test to facial recognition. Who programmed the system, who decided it was ready, and who verified that it could accurately distinguish darker-skinned faces? NIST has measured demographic differences across many facial-recognition algorithms, including higher false-positive rates for African American faces in some one-to-one systems and especially serious disparities affecting African American women in many one-to-many searches. Performance varies across products, but the underlying demand Jones made remains valid: no agency should treat a vendor’s confidence score as neutral evidence when an error can become an accusation against a real person.Children Need Time to Learn the Work
Near the end of the broadcast, Jones and I turned to New York City’s new classroom policy. The precise action is a one-year moratorium during the 2026–27 school year on student-facing generative AI for children from 2-K through eighth grade. It is not a ban on every algorithm used by the school system. High schools may participate in a limited set of approved pilots while the city evaluates the technology.10
Jones supported the moratorium because early education is supposed to build the ability to read, write, calculate, question, and show how an answer was reached. Her point was not that nobody should ever use AI. She uses it in her own company and described building agents, trackers, and workflows with it. The line she drew was developmental: a tool becomes dangerous when a child can obtain an answer before learning how to form the thought.
The same distinction applies to writers. During the livestream, I raised recent accusations about three Black novelists who had been accused of using AI. AI accusations can become a new way to question Black intellectual authorship, but each case must be examined on its own evidence. A punctuation mark, a detector score, or somebody’s disbelief that a Black writer produced polished work is not proof.
What Governance Would Actually Require
Jones rejects both blind optimism and helpless doom. AI can help a small business build tools, organize information, analyze patterns, and reduce work that would otherwise consume scarce time. The answer is not to pretend those benefits are imaginary. It is to stop treating benefit as a waiver of scrutiny.
Real governance would begin before public deployment. It would document the data, identify foreseeable harms, test performance across affected groups, limit collection and retention, separate development from independent review, and give auditors enough authority to delay or stop release. In employment and policing, it would also provide a meaningful path for a person to challenge a result rather than forcing her to argue with an institution that insists the machine is objective.
Most important, the people nearest the likely harm need decision-making power, not ceremonial representation. If a Black woman can describe exactly how a system will fail but cannot change its design, budget, access rules, or launch date, the institution has collected diversity without accepting governance. It wants the appearance of warning without the inconvenience of being warned.
Closing Argument
The thread running through this conversation is not that Laura Loomer, Jason Black, Workday, Google, Flock, and a school district are the same thing. The footage records Loomer’s and Black’s own words, while the Workday claims remain allegations in active litigation. Flock has audit mechanisms, even as officers’ alleged misuse demonstrates the limits of mechanisms without consistent enforcement, and New York City’s classroom moratorium remains a policy experiment whose results still have to be measured.
What connects them is the transfer of power from a stereotype, to a political weapon, to ordinary culture, and then into data, workplace judgment, product design, and enforcement practice. By the time somebody calls the outcome “technical,” the human choices that produced it have been spread across enough people and systems that everybody can claim the decision came from somewhere else.
Ashley Jones gave us a better standard. Listen to the people who find the danger early, and give them enough authority to do something about it. An institution that praises Black women’s resilience while denying Black women the power to prevent the next injury has merely learned how to budget for our pain.
Support XVOA
Paid subscriptions finance the time it takes to recover transcripts, check legal claims, distinguish allegations from findings, and turn a long livestream into reporting that remains public to everyone. Paid subscribers can also use Put XVOA On It to nominate questions and neglected stories for editorial consideration. They help aim the desk, while the reporting stays public. Become a paid XVOA subscriber.
Thank you to the readers already carrying this work. If another recurring subscription is not right for you, or if you are already paid and want to add direct support, Buy Me a Coffee accepts anything from a small contribution to a substantial one-time gift. Like, comment, share, and Restack this Reader’s Cut so the evidence travels farther than the original broadcast.
Sources
Full XVOA interview with Ashley E. Jones
The broadcast contains Jones’s statements, the brief discussion of Gloria Steinem’s death, and the Laura Loomer and Jason Black clips analyzed in this Reader’s Cut.“On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?”
The 2021 FAccT paper by Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell addresses training-data documentation, bias, environmental costs, and the risks of very large language models.“What Really Happened When Google Ousted Timnit Gebru”
Wired documents the competing accounts of Gebru’s departure and the conflict over the “Stochastic Parrots” paper.EEOC amicus brief in Mobley v. Workday, Inc.
The brief explains the agency’s position on how federal employment-discrimination law may apply to an AI screening vendor; it does not decide the merits of the plaintiff’s allegations.Flock Safety License Plate Reader Policy
Flock describes its access controls, audit records, security measures, and recommended customer auditing.Northwestern University on Moya Bailey and misogynoir
Northwestern explains Bailey’s coinage and definition of the term.Ashley E. Jones, “Two Can Play That Game”
Jones’s essay describes using AI strategically inside an employment market shaped by automated screening.“How rogue officers turned a nationwide camera network into a tool for stalking”
The Washington Post reviewed police and court records involving officers charged with or accused of misusing license-plate-reader and other databases.NIST testimony on facial-recognition technology
NIST summarizes measured demographic differentials in face-recognition algorithms and the consequences of false positives.New York City Public Schools guidance on AI and screen time
The policy establishes the 2026–27 student-facing generative-AI moratorium for 2-K through eighth grade and describes limited high-school pilots.
