I grew up boating so when I could afford a small fishing boat, I bought one. We live five minutes from the marina so it’s an easy outing anytime.

I wish I could tell you it’s been smooth sailing since day one, but sadly it hasn’t.

Four times that sucker broke down out on the water. While we’re never far from shore, it’s not much fun bobbing around in an 18 footer with oil tankers and yachts zipping by.

Turned out water kept getting into the gas tank.

After much trial and error over several months, not to mention hefty service costs, my boat service guy figured out the problem was a defective gas cap. The vent seal broke letting rain in, of which we have no shortage in Vancouver.

He literally took the boat apart pressure testing the gas tank only for it to be a $20 part.

It’s been smooth sailing since then fortunately.

After the fourth break-down incident, the service guy asked “do you know you can sign up for a tow service? They’ll come get you and tow you to your marina. It’s $200 per year.”

I had to laugh. I had no clue. I sure would have liked to know that after the first time we broke down in tanker lane.

Needless to say I signed up two minutes after talking to him. Also needless to say, we’ve not had an incident since paying for the service.

All along there was a good solution to breaking down I didn’t know existed.

Stop not knowing by doing this

This not knowing is rife in our industry given we work with tech. Most tech out there can do fifty billion more things than we know about.

Pre-AI we had to ask Google which. This was okay. Sometimes we’d find the solution. Other times, we wouldn’t.

But now, with AI, one of the smartest things you can do from an execution standpoint is to ask AI if there’s a better way.

Not only can AI search for ready-made solutions, it will also happily tell you it can code the solution for you.

At first, I opted for ready-made. That was until I decided to give coding with AI a shot.

It only took that first vibe code experience to realize if AI can code it for me, that’s almost always the best solution.

Not only will AI code it for you, but it’ll often suggest other approaches. If it doesn’t, ask.

I’m still not the most thoughtful vibe coder. I’m the type of person that likes to see some messy prototype that half-works and then improve it from there. I need a visual version, even if it’s not very good. I can always improve it and by seeing it, I know what to fix.

That’s just me.

If you prefer to figure out every parameter you need up front, do that. Doesn’t matter how you go about it as long as you’re asking.

Just yesterday I built something (custom email signup popups controlled from ​my publishing dashboard​) for a couple of sites I should have built ten eons ago. I didn’t because I was still locked in my old way of “3rd party reliance” thinking. I had forgotten that AI is my tech genie… rub the pot and make a wish.

Another example. I needed a lengthy Google form for collecting user generated content. I was just about to start creating it one painful widget a time when I wondered if AI had a faster option.

Claude to the rescue. Claude told me “Jon, sit back and let me handle this for you.”

Sure enough, Claude cranked out some code totally foreign to me and told me exactly what to do with it in Google Cloud Console. Even gave me a link for where to drop the code.

I did as Claude commanded. Ten seconds later my two-page Google form was done.

Always ask AI first.

Don’t assume your audience knows what you know

There’s another lesson here and that is never assume your audience knows what you know. I’ve made that mistake a million times as well.

I suspect my boat service guy assumed I knew about the tow service because after all, what idiot boat owner doesn’t get a tow service? It’s boating 101.

Simple answer. Yours truly didn’t know.

I’ve been slowly getting more active on Substack. What’s interesting is I see all kinds of people having huge success pretty much saying the same thing over and over.

It’s their message.

They repeat it over and over from different angles.

It’s been distilled down to be as simple as possible.

It’s effective.

I tend to do the opposite.

Complicate things. Assume things. Overthink things.

Had I been one to simplify like others, I’d have checked the gas cap first… right after I paid the $200 tow plan.

Thanks for reading.

Jon Fat Stacks

P.S. If you’re creating online or publishing or consulting or doing anything online, consider building and running it all from ​your own custom AI dashboard​.

One of the dumbest mistakes I made publishing online was quitting my first blog.

I shut ‘er down not because I didn’t like the niche but because it hadn’t shovelled a million bucks into my pocket by month six.

The curse of the newb is impatience.

The reason it made squat wasn’t lack of traffic but my having no clue how to monetize.

Looking back, it was a fantastic niche.

I wasn’t a hobbyist. I was an expert.

B2B audience.

Buyers with budgets.

Extremely email friendly.

Problem was I knew nothing about monetization or email. My idea of a slick email campaign was sending the same Hotmail to ten people.

Had I known then what I know now, I’d still be running that site.

If curious about the niche, it was tech for lawyers. At the time I had spent nights and weekends working with consultants to upgrade a law firm’s tech from Word Docs and folders to slick third party legal case management software.

My second blog ended up a success. But it never came close to the potential of the one I quit.

While figuring out the blogging biz, I stumbled on a book that changed how I thought about the whole thing. Sorry, can’t remember the title or author… just the message.

The book’s big idea: the money is in owning the publication, not writing for it.

I took it to heart.

Built up a team and went to town. I wrote too. Full tilt effort all day, every day, cranking out as much content as humanly possible.

That was the playbook for years. And it worked.

Now?

Don’t need a team.

Don’t need to write until the screen goes blurry.

I used to think scaling meant hiring. These days scaling means removing myself from bottlenecks.

Which leads me to what might be the worst advice:

“Just create more content.”

Not because content doesn’t matter. It does.

But because content volume is not a competitive advantage anymore.

Everyone can crank out 200 articles before lunch now. While at the gym. In a language they don’t speak.

So what replaces “just publish more”?

Two words.

Operational leverage.

Or as I’ve been calling it lately, building like an operator instead of a creator.

Here’s the difference.

A creator trades time for content. Butt in chair. Madly typing. Post. Post. Post. Repeat. Output is tied to hours.

An operator builds the system that produces the content. The system runs. The operator checks up on it. Tweaks it. Goes for a walk.

One scales. One fuels burnout.

A few things I’ve noticed about this change…

I. AI didn’t kill creators. It commoditized them.

When everyone can crank content with a one sentence prompt, content speed isn’t an advantage. The advantage moves to those who have a better system producing the writing.

Quality control. Editorial review. Throat-grabbing headlines. Internal linking for pageviews. Distribution. Higher revenue per 1,000 visitors. Revenue tracking.

That’s the advantage now.

II. Creators obsess over the next article. Operators obsess over the system creating the articles/posts.

One is a job. The other is a business.

That cliché still applies, by the way. Owning the publication still beats writing for it. The only thing that changed is what “owning” looks like.

Back then, owning meant hiring a team.

Now, owning means running the operating system.

III. Improvements compound across the system, not one article at a time.

A creator improves an article. The article gets better.

An operator improves the system. Every article going forward gets better.

That’s leverage you can’t get from writing harder.

IV. Creators chase traffic. Operators chase RPMs.

Adding $2 RPM to existing traffic beats chasing another 10K visits almost every time. Creators obsess over top of funnels. Operators squeeze the entire funnel.

Display ad rates. Affiliate placements. Newsletter monetization. Email click revenue. The rev-stack matters as much as audience size.

V. The goal was never “more content.”

It was time. Freedom. A business that doesn’t need you online 16 hours a day.

Systems deliver that.

Hustle never did.

​Build the machine​, or be the machine. Those are the options now.

Thanks for reading.

Jon
Fatstacksblog.com

I watched the movie Margin Call last night.

It’s about the lead-up to the 2008 financial crisis through the lens of a “fictitious” investment bank in NYC.

FYI movie spoiler alert below.

Said bank is helmed by John Tuld, played by Jeremy Irons.

Interestingly, the CEO of Lehman Bros in 2008 was Richard “Dick” Fuld. CEO of Merrill Lynch was John Thain.

That explains the character name “John Tuld.”

Anyway, a junior analyst, formerly a rocket scientist, discovers something awry in the bank’s asset mix signalling imminent financial disaster.

He shows his boss who shows his boss who calls the CEO at 2 am.

CEO swoops in via chopper from swanky estate in the middle of the night.

Junior Analyst explains the situation to the firm’s partners and executives.

Irons’ character immediately understands the seriousness of the situation. He knows the entire model Wall St. operated on for years is broken and will collapse.

He asks the room “what do we do about it?”

Nobody has an answer.

He then says there are only three ways to succeed on Wall St.

1. Be first.
2. Be smarter.
3. Cheat.

He rules out being smarter because there’s always someone smarter. He claims he doesn’t cheat.

That leaves being first, but how?

He proposes the only way out is to be the first to start selling off all problematic assets.

Not everyone likes this because they’ll knowingly sell worthless junk which will shatter their reputations.

Doesn’t matter… CEO declares they have no choice but to liquidate so that’s what they do.

What the movie doesn’t say is that Tuld’s firm got to liquidate first because they were the ones who spotted the problem first.

Anyway, they liquidate as first movers.

As I watched I couldn’t help consider how “being first” applies to online creation.

Like Tuld, I long ago realized trying to be the smartest person in the room is a losing game because there’s always someone smarter.

Cheating isn’t my thing.

Again, that leaves being first, but how?

What does being first look like when creating content online?

There are obvious methods but not always available such as:

• Be first or an early adopter on new platforms.
• Be first to enter new niches and trends.
• Be first to come up with content angles and types.
• Be first to break a story.

That’s about as far as I got on my own. I decided to see what I missed by asking AI (ChatGPT and Claude):

AI spit out plenty (100s) of suggestions.

Here are the ten best:

1. Be first to notice a cultural shift before it becomes obvious in analytics.
2. Be first to operationalize new technology.
3. Be first to package boring data into emotionally compelling stories.
4. Be first to publish the takeaway from a major research paper, government report, or industry study the day it drops. The window is hours, not days, before everyone else has the same idea.
5. Be first to coin terminology in your space.
6. Be first to interview an expert before they’re well-known. Cheaper, more available, and they remember who showed up early.
7. Be first to publish the counter-take the moment consensus hardens. When “everyone agrees X,” the first credible “actually Y” gets outsized attention.
8. Be first to claim a category-level position by name and stick to it (aka category design).
9. Be first to publish predictions with timestamps, then revisit them publicly. Being on the record, right or wrong, is its own moat.
10. Be first to spot and answer what your audience is searching for but no one is answering yet.

​Being first doesn’t always mean inventing something entirely new. Often it simply means noticing something important before everyone else and doing something about it before it becomes mainstream.

The first person to spot the shift doesn’t just gain information. They gain options.

And in business, options others don’t have are a huge advantage.

Thanks for reading.

Jon

Fatstacksblog.com

Creator of The S.M.A.R.T. COMMAND CENTER Course.

I. Introduction

Scenario: A junior associate at a mid-sized firm runs a contract through an AI reviewer. It flags “standard force majeure language,” gives it a green check, and moves on. Weeks later, the client discovers the clause excluded pandemics, exposing them to millions in liability. The AI wasn’t malicious. It had simply learned from a training set heavy with seller-favorable deals.

Bias is woven into every layer: data, design, deployment. When lawyers rely on AI without understanding it, they risk client harm, professional exposure, and erosion of justice. AI isn’t neutral. It carries our assumptions, our histories, our blind spots. Bias cannot be removed. It can only be refined, revealed, and managed. This article explains what that means for lawyers: how bias arises in legal AI, why it matters under ethics and discovery rules, when advocacy-aligned models may be defensible, and how to supervise and prompt them responsibly.

Thanks for reading! Subscribe for free to receive new posts and support my work.

II. What “Bias” in AI Really Means

A. Definition in Plain English

Bias in AI refers to systematic distortion or skew in predictions or outputs. It’s not a random error. It’s when the system consistently favors one outcome, perspective, or group over another.

B. Bias Is a Reflection, Not a Glitch

Every human carries bias shaped by upbringing, culture, and experience. That’s why two juries can reach different verdicts from identical evidence. AI is no different. It reflects the data fed to it and the design choices made by its creators. Bias in AI isn’t a bug. It’s a mirror held up to us.

C. The Three Layers of Bias

1. Data Bias — The training set may underrepresent certain demographics, jurisdictions, or fact patterns.
2. Algorithmic Bias — The mathematical rules may amplify patterns or suppress nuance.
3. User Bias — Lawyers’ prompts, expectations, and interpretations shape outputs as much as any dataset.

D. The Measurement Challenge

There’s no single way to measure fairness. Calibration is where the same predicted probability equals the same actual outcome across groups, often conflicts with equal error rates, where false positives and false negatives are identical. You can’t satisfy both simultaneously. In criminal risk tools, enforcing calibration can worsen disparities in error rates, and vice versa.[1][2] A vendor may claim “fairness” based on one metric while ignoring grave disparities under another.

Key takeaway: Lawyers must ask vendors which fairness standard they use and which tradeoffs they’ve accepted.

III. How Bias Manifests in Legal AI Tools

A. Legal Research and Case Summaries

AI research tools may overemphasize majority opinions or high-citation cases, marginalizing minority or dissenting reasoning. The result: you miss the outlier precedent that could reshape your argument.

B. Document Review and Contract Analysis

Review systems trained mainly on large commercial agreements may treat one-sided clauses as “standard.” That subtle skew translates into costly blind spots for consumer or plaintiff firms.

C. Litigation Analytics and Risk Tools

In 2016, ProPublica analyzed more than 7,000 defendants in Broward County, Florida, and found that the COMPAS risk-assessment tool used in sentencing and bail — produced unequal error rates between racial groups.[3] For defendants with the same risk score, actual recidivism rates were roughly equal across races. The model was calibrated. Yet Black defendants were twice as likely to be labeled “high risk” when they did not reoffend.

White defendants were more often misclassified “low risk.” That clash between calibration and equal error rates remains one of the thorniest fairness problems in AI. For lawyers, COMPAS shows that even mathematically consistent models can encode social inequity. The same dynamic plays out when discovery tools prioritize certain document patterns over others, embedding hidden assumptions about what matters.

D. Hiring and HR Algorithms

Amazon’s now-defunct recruiting AI revealed a parallel problem in corporate settings. In 2018, Reuters reported that Amazon discontinued an internal hiring algorithm after discovering it downgraded resumes containing female-coded terms such as “women’s chess club.”[4] The model had been trained on ten years of resumes from a male-dominated workforce. It taught itself to replicate that imbalance. No one told it to discriminate. The bias emerged naturally from the data.

Practice implication: Bias doesn’t require intent. Once deployed, even “objective” AI can perpetuate inequality quietly, at scale.

IV. Why Bias in AI Is a Legal Problem

A. Ethical Duties

Under Model Rule 1.1, competence includes technological understanding. Formal Opinion 512 (2024) reminds lawyers to know the risks and benefits of AI tools.[5] Rule 5.3 requires supervision of non-lawyer assistants, which includes technology. Rule 1.6 safeguards client information; careless prompt-sharing could expose confidential data. Rules 3.1 and 3.3 prohibit false statements and require disclosure of adverse authority. AI hallucinations can trigger both.

B. Liability Risk

Judges have already sanctioned lawyers for fabricated citations. Delegating judgment to algorithms doesn’t dilute responsibility. It amplifies it.

Further reading: How Lawyers Can Reduce AI Mistakes in Legal Work

C. Discovery Implications

If AI outputs shaped your litigation strategy, it’s possible in due course that opposing counsel may seek discovery on your methodology, datasets, or model versions. Privilege arguments may hinge on documentation. Keep version logs and prompts as part of your file.

D. Regulatory Oversight

The NIST AI Risk Management Framework (2023) emphasizes transparency and fairness. The EU AI Act (2024) designates legal-decision systems as “high risk.” Lawyers can’t plead ignorance of these standards.

Practice takeaway: Bias isn’t abstract. It creates malpractice exposure, discovery burdens, and regulatory risk.

V. Bias Can’t Be Eliminated, Only Refined

No juror or judge is perfectly neutral, yet the justice system aspires to fairness. AI deserves the same treatment. Neutrality may be impossible, but unexamined bias is inexcusable. Lawyers must measure, document, and correct bias as a matter of competence.

Practice takeaway: Treat bias like precedent. It’s always there, but your job is to interpret and refine it responsibly.

VI. Advocacy-Aligned AI: When Bias Becomes a Feature

A. The Proposition

Law firms now experiment with custom GPTs tuned for plaintiff or defense advocacy. A PlaintiffGPT might emphasize causation and damages. A DefenseGPT might prioritize statutory limits and procedural defenses. The question isn’t whether they can. It’s whether they should.

B. Ethical Analysis

Zealous advocacy allows partiality, not falsehood. A biased model that omits controlling law crosses into misrepresentation. Supervision and documentation remain essential. Advocacy bias cannot become discrimination bias.

C. Client Consent

If an advocacy-tuned model could materially influence a case, client consent may be warranted. Clients should know the model’s role and limitations; particularly if its framing could tilt advice or tone. Transparency builds trust and mitigates ethical risk. Informed consent isn’t just good practice. It’s a shield against claims that AI-driven strategy betrayed client interests.

D. The AI Arms Race

Imagine a near future where both sides deploy advocacy-tuned GPTs in discovery or motion practice, each reinforcing its own worldview. Plaintiff counsel’s model surfaces every sympathetic precedent and frames facts for maximum liability. Defense counsel’s model emphasizes causation breaks, statutory bars, and Daubert challenges.

The result: escalating rhetorical extremes, drafted at machine speed.

Courts may respond by mandating disclosure of AI use, requiring neutral audit modes before filings, or applying Daubert-style admissibility tests to algorithmic reasoning. If the profession doesn’t self-regulate advocacy AI, the judiciary may eventually do so.

Practice takeaway: The question isn’t whether advocacy-aligned tools will proliferate. It’s whether lawyers will govern them before courts impose governance from the bench.

E. Helpful, Not Mandatory

Advocacy-aligned tools aren’t necessary, but they’re quickly becoming integral in high-volume practice areas. Used responsibly with oversight, adverse-authority checks, and version control, they can amplify competence. Without those safeguards, they erode it.

VII. The Lawyer’s Role: Governance and Oversight

A. Treat AI Like a Junior Associate

Trust but verify. Every AI draft requires human review.

B. Vendor Due Diligence

Ask AI software vendors how they detect and mitigate bias, what fairness metrics they monitor, and who oversees model integrity.

Exit strategy: Prepare for vendor drift or discontinuation. Models change. Vendors shut down. If a tool you’ve relied on for months suddenly behaves differently or vanishes, you need archived versions, prompts, and outputs to defend your work. Document everything.

C. Version Control

Models evolve. Behavior shifts with updates. Maintain logs of prompts, model versions, and output changes.

D. Firm Governance

Adopt an AI policy outlining usage, confidentiality, review standards, and escalation paths. Periodically audit outputs for bias.

E. Judicial AI Adoption

Courts are starting to rely on AI for administrative tasks but eventually that will expand for case management, analysis, and even the drafting of judicial reasons. While no court of record currently delegates decision-making to AI, early signs suggest the judiciary is already experimenting at the margins. Judges, clerks, and administrative staff are beginning to use generative tools informally for scheduling, summarization, and research support — often without explicit authorization or oversight.

In October 2025, Reuters reported that a U.S. senator had questioned federal judges about whether AI had been used to assist in withdrawn court rulings.[7] The example linked above is a warning to the profession: judicial personnel will use AI, whether sanctioned or not. That reality makes governance, not prohibition, the more prudent path.

Governing bodies must now decide whether to formally sanction AI use at the judicial level — and if so, how to supervise it. Transparent policies and audit trails could prevent the quiet proliferation of unsupervised tools within the court system. It is arguably better for AI to be used safely and openly than for judges to go rogue in the shadows of innovation.

Key takeaway: Lawyers should anticipate, not merely react to, judicial AI. The profession must help set the standards for explainability and oversight before courts adopt these tools at scale. Efficiency without transparency is not progress — it is abdication.

VIII. Practical Guidance

It’s all well and good to discuss these issues in theory, but the real value lies in guidance for putting it into practice. Lawyers can’t afford to treat AI bias as an abstract ethical dilemma. It’s a day-to-day operational issue, one that affects vendor selection, client communication, and even how you draft a simple prompt.

It’s easy to rely on and trust AI software vendors, and in many cases, it’s reasonable to do so. But that doesn’t mean attorneys or firms should blindly sign up and deploy. Every firm has an obligation to understand the bias (and alignment) baked into the tools they use. You wouldn’t outsource legal research to an associate you’ve never met or supervised. The same standard applies here.

A. AI Software Vendor Questions

Ask vendors:

• What data trained the model, and how is bias tested?
• Which fairness metrics do you use?
• How often are audits performed, and are they independent?
• Can you trace reasoning and cite sources?
• How do you communicate model updates?
• Who oversees bias internally?
• For your practice area, are datasets balanced by plaintiff/defense or buyer/seller orientation?

Further reading: How to choose the best AI software for your firm.

B. Prompting for Fairness and Advocacy

Prompts are the lawyer’s new form of instruction. Just as phrasing can sway a witness or frame a cross-examination, the way you phrase an AI query determines what perspective the system adopts. Mastering this skill means understanding when to ask for balance and when to advocate.

Neutral / Research Mode:

“List both majority and dissenting opinions on [issue].”
“Summarize circuit splits and explain the reasoning.”
“Identify what opposing counsel might argue.”

Advocacy Mode:

“Draft arguments supporting the plaintiff assuming favorable facts.”
“Include adverse controlling cases the court would expect.”
“Write a defense motion emphasizing procedural issues; verify citations.”

Two-pass workflow:

Many firms are adopting a two-pass workflow: one system generates the initial draft, and another independently reviews it for accuracy and bias. It’s the digital equivalent of a second set of eyes.

• Generate in advocacy mode.
• Run a neutral review prompt to surface omissions or bias.

Practice implication: Precision in prompting shapes fairness and credibility.

IX. Case Studies

A. COMPAS (Criminal Sentencing)

In 2016, ProPublica found that the COMPAS risk-assessment tool in Broward County, Florida, was statistically calibrated but racially skewed.[3] Defendants with equal scores reoffended at similar rates. Yet Black defendants were twice as likely to be labeled “high risk” without reoffending. White defendants were more often misclassified “low risk.” The takeaway: technical accuracy does not guarantee justice. Fairness is a choice, not an output.

B. Amazon Recruiting AI (Employment)

Amazon’s internal recruiting model downgraded resumes containing female-coded terms like “women’s chess club.”[4] Trained on male-dominated historical data, it learned patterns that excluded women without explicit instruction. For employment lawyers, that case is a reminder that bias can emerge without intent. Intent is irrelevant when outcomes are discriminatory.

C. Predictive Policing (Law Enforcement)

Predictive policing models often amplify disparities. Over-policed neighborhoods yield more crime data, which reinforces future patrols.[6] The bias becomes cyclical. Lesson: Bias propagates through feedback loops. Unchecked, it becomes precedent masquerading as probability.

X. The Path Forward

Lawyers must embed algorithmic literacy into professional competence. Bias will always exist, but it can be understood, monitored, and corrected. The rule should be simple: Explain or abstain. If you can’t explain a model’s reasoning, don’t rely on it.

XI. Conclusion

When I began studying legal AI, I believed bias was a technical flaw that could be coded away. I was wrong. Bias is a reflection of us — our data, our priorities, our judgments. Eliminating it is impossible. Refining it is our duty. The tools we build mirror the system we serve. Lawyers must ensure that reflection bends toward fairness — and that justice never cedes to the algorithm.

My Take

I’m going out on a limb here because it’s not talked about much. Lawyers have an ethical obligation to use biased AI when that bias mirrors zealous advocacy for their clients. I’m not suggesting manipulation, suppression of evidence, or anything unethical. I’m talking about aligning AI’s analytical leanings with a lawyer’s legitimate role of presenting a client’s case in the best possible light. A pro-plaintiff or pro-defendant bias is not malpractice; it is strategy.

Firms that invest in custom-built AI tuned to their advocacy style may gain a real edge. These systems can frame arguments, interpret evidence, and prioritize precedents through the lens of client advantage. Neutral AI will not do this as effectively.

And it will not stop there. Smart firms will also deploy mirror models, AI trained to think like the opposition, to anticipate counterarguments and simulate adversarial reasoning. This adds cost and complexity, but nuanced advocacy-aligned AI will soon become a competitive moat for firms that take representation seriously.

For now, only customized systems can achieve this level of alignment. In time, the major legal AI platforms will introduce similar advocacy modes. When that happens, the firms that already understand how to wield bias ethically will be the ones ahead of the curve.

Originally published on JurisIQ.io.

References

[1] Alexandra Chouldechova, “Fair Prediction with Disparate Impact: A Study of Bias in Recidivism Prediction Instruments,” Big Data 5(2): 153–163 (2017). Published by Mary Ann Liebert, Inc. https://doi.org/10.1089/big.2016.0047

[2] Pleiss et al., “On Fairness and Calibration,” Advances in Neural Information Processing Systems 30 (2017). https://papers.nips.cc/paper/7151-on-fairness-and-calibration.pdf

[3] Julia Angwin et al., “Machine Bias,” ProPublica (May 23 2016). https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing

[4] Jeffrey Dastin, “Amazon Scraps Secret AI Recruiting Tool That Showed Bias Against Women,” Reuters (Oct 10 2018). https://www.reuters.com/article/amazon-com-jobs-automation-insight-idUSKCN1MK08G

[5] American Bar Association, Formal Opinion 512: Generative Artificial Intelligence Tools (July 2024). https://www.americanbar.org/content/dam/aba/administrative/professional_responsibility/aba-formal-opinion-512.pdf

[6] Kristian Lum & William Isaac, “To Predict and Serve?” Significance 13(5): 14–19 (2016). Published by Wiley on behalf of the Royal Statistical Society. https://doi.org/10.1111/j.1740-9713.2016.00960.x

[7] David Thomas, “U.S. Senator Asks Judges if They Used AI in Withdrawn Court Rulings,” Reuters (Oct. 6, 2025). https://www.reuters.com/legal/government/us-senator-asks-judges-if-they-used-ai-withdrawn-court-rulings-2025-10-06/

Thanks for reading! Subscribe for free to receive new posts and support my work.