Bill Acton's Embodiment Blog: When AI gets it (and you) wrong: Three recent cases

Wednesday, October 14, 2026

When AI gets it (and you) wrong: Three recent cases—and the fight to put things right

An AI mistake can be amusing when it invents a wonky book title. It is something else when it helps put you in jail, bars you from entering a place of business or steals your identity.

The problem does not necessarily begin with AI “going rogue.” Machines sometimes produce questionable answers. A human may accept one uncritically, too readily, and an institution then acts on it. Some poor soul on the receiving end is then left with the often monumental task of trying to prove what should never have been in doubt in the first place. Sound familiar?

Recent legal developments illustrate the danger. A Tennessee grandmother alleges that a flawed facial-recognition identification and inadequate police work left her detained for nearly six months. A Nevada man’s case describes an officer trusting a casino’s facial-recognition result over his valid identification. A newly filed lawsuit against OpenAI raises a different question: what happens when children trust a chatbot with sensitive personal information, but the safeguards allegedly fail?

These are not identical cases. Nor have the allegations all been established in court. But together they bring the discussion about AI down to a very personal level: When an AI-assisted system gets your life or identity wrong, what recourse do you have?

A grandmother’s quiet life: evidence of her innocence

In September, 2026, Angela Lipps filed a federal civil-rights lawsuit against Wells Fargo - North Dakota, and a detective. The Tennessee grandmother seeks $10 million, alleging that a flawed facial-recognition identification and inadequate investigation led to nearly six months of detention over bank fraud that she did not commit.
Her bank records and other sources finally showed her living in Tennessee--while the crimes occurred in North Dakota. The lawsuit alleges that police treated the facial-recognition lead as positive identification--and failed to give proper weight to other evidence of her innocence.
The issue is not merely whether a machine produced a mistaken identification. It is whether authorities tested that lead adequately before making a life-changing decision. (Lipps’s damages claim remains unresolved.)

Sources:

[foxnews](https://www.foxnews.com/us/grandmother-sues-alleged-ai-facial-recognition-error-led-months-jail-experts-sound-alarm)
[wcyb](https://wcyb.com/resources/pdf/61ca4579-1927-48a0-be1c-b05675b0568e-AngelaLippsLawsuit.pdf)

When a driver’s license loses an argument with a computer

In September 2023, facial-recognition technology at Reno’s Peppermill Casino identified Jason Killinger as a possible match for someone barred from the property. According to the ACLU, he presented a valid commercial driver’s license—but the officer trusted the computer and concluded, without supporting evidence, that the license must be fraudulent. Fingerprints later confirmed Killinger’s identity; the technology was wrong.
The casino previously settled a separate lawsuit for an undisclosed amount, while his civil-rights case against the officer and city continues. Neither the settlement nor the groups’ legal arguments establish liability, however.

Source:

[aclu](https://www.aclu.org/cases/killinger-v-jager)

What children tell a chatbot

A nationwide class action filed against OpenAI on October 13 alleges that ChatGPT collected children’s sensitive information without parental consent and that OpenAI used it to train models and build profiles. (The proposed class of the suit has not yet been certified.)
According to Bloomberg Law, one plaintiff allegedly started using ChatGPT at 12. The complaint says conversations continued after she disclosed her age; she later shared selfies and discussed her appearance, insecurities, diet, and body. The plaintiffs allege that the product encouraged trust and disclosure without adequate safeguards.
These are only allegations, not findings, of course. Unlike the arrest case, this dispute concerns consent, privacy, and product design. But it raises a similarly personal question: who is responsible when trust in an AI system allegedly leads to harm and how can one respond?

Sources: [news.bloomberglaw](https://news.bloomberglaw.com/us-law-week/openai-accused-of-taking-minors-data-without-parental-consent)

Correcting the error is not repairing the harm

The connecting issue is "accountability." Better software may prevent future mistakes. It does not resolve a detained person’s damages claim or determine what should happen to information allegedly collected improperly. Those questions involve evidence, rights, and legal responsibility.

Source: [news.bloomberglaw](https://news.bloomberglaw.com/us-law-week/openai-accused-of-taking-minors-data-without-parental-consent)

Access to legal advice can provide a starting point. LegalShield, for example, describes benefits including consultation, document review, and qualifying letters or calls. Extensive representation may require separate arrangements; coverage depends on the plan and terms. Civil trial defense is not the same as funding a lawsuit you bring. The promise is access to help—not guaranteed recovery.

Source: [legalshield](https://www.williamacton.legalshieldassociate.com/)

The troubling possibility these cases raise is not that AI has become all-powerful . . . yet. It is that we can give its outputs more authority than deserved. Should that happen, an ordinary individual may face an extraordinary task, getting an institution to reconsider—and then seeking a remedy for what its decision has already cost.

No comments:

What is haptic pronunciation teaching?

Basically, it is using gesture, touch and movement to teach pronunciation of English (or any language). What makes this method unique is that it is based on controlled and systematic use of gesture in doing that. (Often the problem with gesture use in language learning is that it can be "crazy" or random or intimidating or inconsistent, or not well-developed and tied to clear outcomes.)

The earlier "Essential Haptic-integrated English Pronunciation (EHIEP) approach," now KINETIK, does a better job of enhancing presentation, modelling, feedback and correction. The method is simple and easy to follow. Once a 30-minute module is finished, teacher and students have a set of techniques to use in class anytime a pronunciation problem or opportunity comes up!

Haptic Preliminaries

I often begin workshops or papers with the comment that in about 40 years in the field I have had just one idea: that the systematic use of body movement is essential to effective and efficient pronunciation instruction. That has never been more relevant than today, with the general deemphasis on pronunciation and introduction of technology into the field. From both perspectives, the "haptic" perspective approach developed here offers great promise. I can at least promise you a "moving" experience!

What is the KINETIK Method?

KINETIK is the name of the new v6.0 haptic pronunciation teaching system. About six years earlier we were using the acronym, "Haptic Integrated Pronunciation for Other/Extended Circle English Speakers" (HIPOECES)--still there in the url of this blog!

Haptic-integrated clinical pronunciation is:

A different way to learn pronunciation, based in part on Arthur Lessac’s notion of “Training the body first.” It is also based on research in the use of movement and touch in learning and training. It:

Looks somewhat like a combination of aerobics, sign language and Taichi

Provides a basic foundation for continued, self-directed pronunciation learning and classroom instruction

Designed for use by relatively untrained teachers but appropriate for all teachers (and learners) of all levels

Focuses on pronunciation used in conversation (not all words in English)

Is a highly “brain-and-body-friendly” system that promotes efficient learning of integrated tasks in general

Designed to promote integrated classroom pronunciation instruction and enable integration of conscious pronunciation work into spontaneous speaking.

Introduction

This blog was formerly named: HIPoeces (pronounced 'hI-pƏ-cIs), basically an approach to teaching English pronunciation to speakers of "outer and expanding" circle Englishes (including that population commonly termed: non-native speakers) but also compatible with the use of local models other than simply Standard" American" or RP. Using Kachru's "Inner, expanding and outer circles" model to characterize the breadth of Englishes and speakers of those, HIPoeces (now HICP) systems focus on developing haptically-centered pedagogical models and methods which can be adapted for teaching the pronunciation of various Englishes as they are spoken and taught by "owners" of those varieties.

In theory, the underlying principles can be applied to teaching the sound system of any language. Most language instructors intuitively use a great deal of movement and tactile anchoring in the classroom. The essential question is how to use that resource systematically, especially so that those learners who are less "haptic" in nature are also "moved" by the method.

Simply put, the HICP model/method is based on the idea of speaking along with arm and hand movement, not all that different from that used in sign languages. Most movement patterns are taken from either typical classroom pronunciation instruction practice or from related fields, such as psychotherapy or sports.

Most blog posts link to research that relates to the general HICP model, that is studies dealing with the role of gesture, touch, movement, learning modalities, learning in output, use of the visual field in learning, and experiential learning.

DEFINITIONS:

"Haptic" techniques- procedures that involve both movement and touch (Obvious examples include clapping hands, tapping feet, or holding an object and shaking it to the rhythm of a poem.)

"HICP techniques" - procedures that are haptic and also explicitly coordinated with phonological structures and/or discourse prosody (For example, clapping hands on primary stressed syllables in multi-syllable words, or doing a dance step on key words in conversational turns.)

"Haptic integration"- The haptic dimension probably functions as much to anchor visual and auditory material as it does to assist the learner in getting the "felt sense" of the sound or sound process. Lack of integration into spontaneous speech being such a serious potential shortcoming of instruction, haptic engagement seems a promising bridge to more accurate fluency.

"Visual field" - in HICP extensive use is made of the visual field in front of the learner. (For example, a modified version of the IPA vowel chart is anchored haptically in the imagination of the learner, directly in front of the upper body, and used for several purposes throughout the program.)

Wednesday, October 14, 2026

When AI gets it (and you) wrong: Three recent cases—and the fight to put things right

No comments:

Post a Comment