AI Agents And Kids
Copyright Alexandra Koch/Pixabay

AI Agents And Kids

Updated June 3, 2024

A revolution is just beginning. As AI agents sweep into our lives, they'll also rapidly sweep into our kids lives as well. Yet, it's not simply applying AI and tech to kids. It brings with it all sorts of new challenges including legal minors, content, privacy, authorization agreements, what a child's AI agent can and can't do etc.

I've spent the last 8 years slowly working my way through this. I hope your eyes are opened wide from this article! It's out of the box thinking for our out of the box times. So, if you'd like to learn more, read on...

Note: This article is also available as a PDF.

Story Line This Article Uses

Mary and John Doe are children of Dr. Jane Doe.:

  • Mary is autistic and has a brilliant mind. Thus, Jane leverages a physical bot (like QT robot) and an AI agent learning assistant (AssistBot) for Mary
  • John was borne with a medical condition. Jane leverages IoT devices, and a medical AI agent for John (MedBot), from when he's born, until he's an adult, to continuously monitor his medical condition
  • From when they're toddlers, both children have a DLT (Digital Learning Twin), which continuously updates their own IEP (Individualized Learning Twin) fed with data from their personal AI agents and bots
  • All the learning data for each child is stored in their LDV (Learning Data Vault) database, which Jane controls until they come of legal age
  • As they grow up, they also have AI agents they use in games, and when they're teenagers, different AI agents to assist them with banking (BankBot), shopping (Shopbot), etc.
  • When both kids enter school, Jane gives her permission for the school to use their DLT, IEP and LDV. Mary is able to do university level courses when she's in elementary school. Mary takes her AssistBot and QT Robot to school with her.
  • Mary and John are taught by Sally GoodTeacher and her teaching assistant bots, PattyBot and BobBot in an AI/AR/VR learning environment

As an aside, to see what's coming at us re AI agents, skim these articles:

Challenge #1 - John & Mary's Legal and Digital Identities From When They're Born

Problem Statement:

Today, on the planet, there's a crappy legal identity framework. How can I say this? There are:

  • No global data standards for legal identity
  • No ability to query all CRVS (Civil Registration Vital Statistics) registries to confirm an identity
  • No ability to biometrically tie the person holding the CRVS piece of paper (e.g. a birth certificate) to the legal identity stored within a CRVS data system
  • No ability, where risk warrants it, to register legal digital identities against the underlying legal physical identity (like Mary and John's AI agents)
  • No ability to instantly tie Mary and John's legal identity to that of their mother, Jane Doe, which works physically and digitally, locally and globally
  • No ability to tie legal authorization rights to a person e.g. Jane's ability to control what AssistBot, MedBot, BankBot and ShopBot can do and can't do with Mary and John's sensitive data
  • No ability for Jane, Mary and John to easily, instantly, anonymously prove they're a human (not a bot) and if they're above or below age of consent, which works physically and digitally, locally and globally
  • No central data store where Jane, Mary and John can store all legal identity and learning consents given on their behalf or, which they've given, from cradle to grave
  • No ability, based on risk, to register legal identities for AI systems and bots

Solution Strategy:

I've created a new legal identity framework for humans, AI systems, AI agents and bots.

For Humans:

  • At birth the person's fingerprints are obtained and, at a later date, when they can keep their eyes open, their iris scans are obtained. This data is fed into the CRVS birth registry system (for those who don't have fingers or eyes other biometrics will be used)
  • Their legal identity relationship (e.g. parent/child, legal guardian/child, etc.) is cryptographically cross linked within the CRVS data base (e.g. Jane's legal identity is cross-linked with Mary and John)
  • The data is automatically pushed out of the CRVS to each person's SOLICT (Source of Legal Identity & Credential Truth) database. Each person (or their parent/legal guardian) manages this
  • The person's SOLICT in turn pushes out the legal identity information to one of four types of LSSI (Legal Self-Sovereign Identity) devices:
  • A physical smart identity card
  • A digital legal identity application
  • A physical wristband, biometrically tied to the person
  • A chip implanted into the person
  • The person uses their LSSI devices to manage their consents to release portions of their legal identity and/or learning information
  • The person can, hypothetically. in turn leverage their AI leveraged PIAM (Personal Identity Access Management) service to preset it to release portions of their legal identity and/or learning data to others
  • Note - within each person's PIAM it's also cryptographically cross linked to other people's with whom they have a legal identity relationship (e.g. Jane's SOLICT is cross linked with the SOLICTs of Mary and John)
  • Also note that a person who has a legal identity relationship with another can hypothetically manage the person's /entity's legal identity and learning data via their PIAM. For example, Jane can control/manage Mary and John's legal identity and learning data via her PIAM, as well as their associated AI agents (like MedBot, AssistBot, BankBot and ShopBot), as well as physical bots like QT Robot
  • All the above works locally and globally, physically and digitally
  • Skim this story to see how Jane Doe leverages her legal identity to manage her daily life
  • Then skim this doc to see the high level architecture

Challenge #2 - Where Risk Warrants It, AI Systems and Bots Require Legal Identities

Problem Statement:

An AI system, in one jurisdiction on the planet, can create increasingly smart digital bots at speeds of thousands or more per second. In the next second, they're operating in all other jurisdictions on the planet. Today, on the planet, there isn't a:

  • Local/global legal identity framework
  • Which can securely work, at transactional speeds, writing unique legal identifiers into the entity's source code

Thus, there isn't a way for citizens, companies, enterprises and different levels of government to instantly determine entity friend from foe:

  • How can Mary or John be sure it's their AssistBot, MedBot, BankBot, ShopBot etc. they're interacting with?
  • How can a school, bank or retailer be sure it's Jane's, Mary's or John's bots they're interacting with?

There's also no ability today on the planet to instantly determine if it's an AI system or bot you're interacting with:

  • How can Jane, Mary or John determine if it's a bot they're interacting with?

Solution Strategy:

I've created a new legal identity framework for AI systems, AI agents and bots. It is similar to the one used for humans:

  • Each entity is registered within the CRVS system
  • Where risk warrants it, legal identity relationships are created between different entities (e.g. AI System 12345 can create one or millions of digital bots with which it has a legal identity relationship with like BotABCDE)
  • Each entity is given their own SOLICT
  • With their own LSSI device - written to their source code
  • With their PIAM to use to manage their identity relationship consents
  • The entity can instantly use their LSSI device and/or PIAM to prove, anonymously, they're a bot and not a human

Note: While architecting the above I was concerned about:

  • How we'd write, at transactional speeds, unique legal identifiers to their source code? I suspect, but don't know it will require a new programming language
  • The security risks of giving digital entities their own SOLICT's, LSSI device an PIAM? I can easily see malicious people leveraging this to create denial of service type attacks, etc.
  • The lifespan of some types of digital entities? It could range from seconds to decades or centuries.

Thus in the 500 plus page cost centre doc, it allocated lots of funds to rapidly bear down on this i.e., I'm not saying it all is easy to do.

Skim this doc to see the high level architecture.

Challenge #3 - Legal Identity is Frequently Managed At State/Provincial Levels

Problem Statement:

They manage legal identity via their own laws and regulations. They will resist change if it means they lose control of this. Thus, years ago, I realized this was a political hill I could easily die upon if I didn't architect a solution framework still leaving them in control.

So how can Jane, Mary and John, who live in Jurisdiction X have a legal identity framework that works locally and globally?

Solution Strategy:

The architecture still leaves each local state/jurisdiction in control. However, it plugs them into a new global legal identity framework working at what I call "warp speed". Skim this doc aimed at leaders of states/provinces.

Challenge #4- Evil Inc.'s & Malicious States Rapidly Creating New Attack Vectors Against The New Legal Identity Architecture

Problem Statement:

Add to this mess the fact that the Evil Inc.'s of the planet will leverage this curve, to create new attack vectors, DAILY, against the legal identity framework for humans, AI, AI agents, and bots. I realized long ago, most local jurisdictions DON'T have the resources, budgets or expertise to constantly defend their legal identity frameworks against these types of attacks. Thus, I knew any architecture I created MUST address this. Without this, there won't be trust in identities.

Solution Strategy:

The architecture creates two new independent, extremely well funded non-profits. Part of their responsibilities is to do 24x7x365 threat analysis. I strongly suggest readers skim to the sections titled "Which Led Me To Create In The Architecture Two New, Global, Independent Non-Profits" and "How Are They Funded?" in “State/Provincial Leaders -? A Major Problem Is Heading Your Way”.

Thus, Jane, Mary and John might receive messages to update their human legal identities or those of their AI agents, etc. within hours for a very high threat. This is now the architecture brings current industry best practices to the world of legal identity and learning.

Challenge #5 - How Can Jane Effectively Manage Her Kids & Their AI Agents/Bots In Real Time?

Problem Statement:

Like many parents, Jane has a busy work life. Thus, as her kids, their AI agents and bots, interact with an increasing plethora of entities both human and digital, how can she manage this in real time? She's very concerned about their safety, their data and privacy.

Solution Strategy:

The architecture allows Jane to do most of this automatically via her PIAM. To see an example, skim to page 35 in “Cost Centres Rethinking Legal Identity Learning Vision”. It shows a pic of Jane managing her son and his AssistBot with the local school and their LMS (Learning Management System). As conditions change, hypothetically, Jane could preset conditions to automatically approve or, have her notified, to see if she'll approve or not.

Challenge #6 - How Can Jane Effectively Manage A Transition Of Granting More Authorization Rights To Mary and John As They Become Teenagers?

Problem Statement:

As Mary and John become teenagers, Jane wants to assign them increasing responsibilities to their AssistBots, BankBots, ShopBots, etc. within authorization rules she determines.

  • How can she do this?
  • How can banks, retailers, etc. manage this?

Solution Strategy:

The architecture leverages TODA files which, hypothetically, could be authorization rights. Let's use the bank as an example:

  • Acme Bank Inc. creates a legal agreement agreement with Jane specifying a range of authorization rights she can grant to Mary and John. Let's say it's between abilities to withdraw funds of between $1-200.
  • They then issue a TODA file to Jane, allowing her authorization rights to grant Mary and John limited authorization withdrawl rights of between $1-200
  • Jane might start with granting a limited withdrawl right of $1-50 to both Mary and John's BankAgent. Her PIAM issues a TODA file to both Mary and John's BankAgent stating the withdrawl abilities are between $1-50.
  • Mary or John's BankAgent then logs onto Acme Bank Inc.s website. Their identities are instantly verified by Acme Bank Inc. as well as confirming their legal identity relationship is parent/child, via their SOLICT's. Acme might also make a quick trip to the CRVS to ensure the CRVS's digital signature is still valid
  • Assuming so, system then checks the authorization rules for the withdrawl. Assuming it passes, they'll then likely run metadata to ensure the withdrawl fits in with their past history, etc.
  • Assuming this passes, Acme Bank Inc. will then grant the withdrawl of funds from the bank account

To learn about TODA skim “TODA, EMS, Graphs – New Enterprise Architectural Tools For a New Age”.

Challenge #7 - How Can A Person Or Entity Prove Their Credentials

Problem Statement:

As we create an increasingly integrated physical and digital world, how can a person or an entity be able to instantly prove, either physically or digitally, locally or globally, their credentials:

  • How can Jane be sure Sally Goodteacher, PattyBot and BobBot have the appropriate teaching credentials?
  • How can Mary or John prove their courses and/or degrees, etc. they've obtained?
  • How can credential issuing bodies still keep control of their credential granting standards, etc. while plugging into a new local/global credential issuing framework?

Solution Strategy:

  • The architecture allows each credential issuing body on the planet to still keep control of their credential standards, granting processes etc.
  • HOWEVER, when the credential body wants to issue the actual credential to the person or entity, they adhere to the standards and processes issued by the new legal identity, independent, very well-funded non-profit
  • They then issue the credential to the person or entity's SOLICT
  • Thus as Mary or John complete courses in school, obtain degrees or professional or trade certifications, etc. all of this is written by the credential issuing body to their SOLICT digitally signing them.
  • The same process applies to credentials granted to their AI agents
  • Mary or John are now in control of their credentials
  • The same process applies to AI systems, bots, etc.

Skim “Verifiable Credentials For Humans and AI Systems/Bots”.

Challenge #8 - How Can The Solutions Work For All People Regardless Of Their Abilities Or Disabilities?

Problem Statement:

People who have different abilities, learning styles or disabilities are often constantly penalized when the solution frameworks being offered to them don't work for them.

  • How can the new legal identity and learning framework work for all people?
  • How can it work for someone like Mary?

Solution Strategy:

I recently rewrote the entire architecture addressing this. I architected co-design into most of the architectures cost centre and business processes. I wanted to leave no-one behind on the planet regardless of their abilities or disabilities.

I strongly suggest readers skim:

To Make All The Above Magic Work

Requires national and state/provincial leaders to recognize they must fund, design and deploy the new legal identity architecture. To see my messages to the government and industry leaders, skim:

Costs To Fund, Design & Deploy

Price tag? Between $21-35 billion over 3 years. Skim:

Why Will A Government Fund This?

  1. National Security
  2. Give Your industries A Significant Competitive Edge
  3. Provide Your Citizens New Privacy Toolkits Addressing Privacy
  4. Rethink Learning & Training
  5. Works For All Your Citizens Regardless Of Their Abilities/Disabilities
  6. Architecture Cost Centres Designed With Rapid Innovation In Mind

Skim “Why Should Your Government Fund The Architectures?” to learn more.

Meanwhile Within An Enterprise What Can You Do?

Rethink your identity and authorization architecture - Skim? these four enterprise architecture docs :

Then skim “AI Agent Authorization - Identity, Graphs & Architecture”.

Next skim “Zero Trust On Steroids! Rethinking Security Models For Citizens And Enterprises In The Age of AI Agents And Tech”.

Finally, consider using children age verification and prvacy services provided by companies like Privo.

Summary - Our Kids Are Entering A Major Paradigm Shift

Where our old ways won't work well anymore. Thus, it requires out of the box thinking for our out of the box times. That's what the architecture delivers. Our kids deserve the benefits from the architectures.

If you'd like to chat, contact me.

PS I'm Not Always The Sharpest Knife In The Drawer!

I've led several groundbreaking identity projects. Often, on the diverse teams I've assembled to design, deploy and maintain, a person might effectively say, "Guy, YFOS (you're full of shit)! Didn't you know about Y when you wrote X? Duh!'

I love this. Why? It's by going through this process that bad ideas are ditched and good ones are modified by LOTS of testing, POC's (proof of concepts) and pilots before rolling it out into Production. Thus, I fully expect the architectures I've proposed to change over time, as brighter sparks than me come up with better ideas.

What I've done is laid out at the 100,000 foot level a high level architecture, costs, etc. which can now be discussed, criticized and likely changed over time.

About Guy Huntington

I'm an identity trailblazing problem solver. My past clients include Boeing, Capital One and the Government of Alberta's Digital Citizen Identity & Authentication project. Many of my past projects were leading edge at the time in the identity/security space. I've spent the last eight years working my way through creating a new legal identity architecture and leveraging this to then rethink learning.

I've also done a lot in education as a volunteer over my lifetime.?This included chairing my school district's technology committee in the 90's - which resulted in wiring most of the schools with optic fiber, behind building a technology leveraged school, and past president of Skills Canada BC and Skills Canada.

I do short term consulting for Boards, C-suites and Governments, assisting them in readying themselves for the arrival of AI systems, bots and AI leveraged, smart digital identities of humans.

I've written LOTS about the change coming. Skim the?over 100 LinkedIn articles?I've written,?or my webpage?with lots of papers.

Quotes I REALLY LIKE!!!!!!:

  • We cannot solve our problems with the same thinking we used when we created them” – Albert Einstein
  • “Change is hard at first, messy in the middle and gorgeous at the end.” – Robin Sharma
  • “Change is the law of life. And those who look only to the past or present are certain to miss the future” – John F. Kennedy

Reference Links:

An Identity Day in The Life:

My Message To Government & Industry Leaders:

National Security:

Rethinking Legal Identity, Credentials & Learning:

Learning Vision:

Creativity:

AI Agents:

Architecture:

AI/Human Legal Identity/Learning Cost References

AI Leveraged, Smart Digital Identities of Humans:

CISO's:

Companies, C-Suites and Boards:

Legal Identity & TODA:

Enterprise Articles:

Rethinking Enterprise Architecture In The Age of AI:

LLC's & AI:

Challenges With AI:

New Security Model:

DAO:

Kids:

Sex:

Schools:

Biometrics:

Legal Identity:

Identity, Death, Laws & Processes:

Open Source:

Notaries:

Climate Change, Migration & Legal Identity:

"Human Migration, Physical and Digital Legal Identity - A Thought Paper

Fraud/Crime:

Behavioral Marketing:

AI Systems and Bots:

Contract Law:

Insurance:

Health:

AI/AR/VR Metaverse Type Environments:

SOLICT:

EMP/HEMP Data Centre Protection:

Climate:

A 100,000-Foot Level Summary Of Legal Human Identity

  • Each person when they’re born has their legal identity data plus their forensic biometrics (fingerprints, and later when they can keep their eyes open – their iris) entered into a new age CRVS system (Civil Registration Vital Statistics - birth, name/gender change, marriage/divorce and death registry) with data standards
  • The CRVS writes to an external database, per single person, the identity data plus their forensic biometrics called a SOLICT “Source of Legal Identity & Credential Truth).?The person now controls this
  • As well, the CRVS also writes to the SOLICT legal identity relationships e.g. child/parent, cryptographically linking the SOLICTs.?So Jane Doe and her son John will have cryptographic digitally signed links showing their parent/child.?The same methodology can be used for power of attorney/person, executor of estate/deceased, etc.
  • The SOLICT in turn then pushes out the information to four different types of LSSI Devices “Legal Self-Sovereign Identity”; physical ID card, digital legal identity app, biometrically tied physical wristband containing identity information or a chip inserted into each person
  • The person is now able, with their consent, to release legal identity information about themselves.?This ranges from being able to legally, anonymously prove they’re a human (and not a bot), above or below age of consent, Covid vaccinated, etc.?It also means they can, at their discretion, release portions of their identity like gender, first name, legal name, address, etc.
  • NOTE: All consents granted by the person are stored in their SOLICT
  • Consent management for each person will be managed by their PIAM “Personal Identity Access Management) system.?This is AI leveraged, allowing the person, at their discretion, to automatically create consent legal agreements on the fly
  • It works both locally and globally, physically and digitally anywhere on the planet
  • AI systems/bots are also registered, where risk requires it, in the new age CRVS system
  • Governance and continual threat assessment, is done by a new, global, independent, non-profit funded by a very small charge per CRVS event to a jurisdiction to a maximum yearly amount.

A 100,000-Foot Level Summary Of The Learning Vision:

  • When the learner is a toddler, with their parents’ consent, they’ll be assessed by a physical bot for their learning abilities.?This will include sight, sound, hearing and smell, as well as hand-eye coordination, how they work or don’t work with others, learning abilities, all leveraging biometric and behavioral data
  • All consents given on behalf of the learner or, later in the learner’s life by the learner themselves, are stored in the learner’s SOLICT “Source of Legal Identity & Credential Truth
  • This is fed into a DLT “Digital Learning Twin”, which is created and legally bound to the learner
  • The DLT the produces its first IEP “Individualized Education Plan”, for the learner
  • The parents take home with them a learning assistant bot to assist the learner, each day, in learning.?The bot updates the DLT, which in turn continually refines the learner’s IEP
  • All learning data from the learner is stored in their LDV “Learner Data Vault”
  • When the learner’s first day of school comes, the parents prove the learner and their identities and legal relationship with the learner, via their LSSI devices (Legal Self-Sovereign Identity)
  • With their consent, they approve how the learner’s identity information will be used not only within the school, but also in AI/AR/VR learning environments
  • As well, the parents give their consent for the learner’s DLT, IEP and learning assistant bot to be used, via their PIAM (Personal Identity Access Management) and the learner’s PIAM
  • The schools LMS “Learning Management System” instantly takes the legal consent agreements, plus the learner’s identity and learning information, and integrates this with the school’s learning systems
  • From the first day, each learner is delivered a customized learning program, continually updated by both human and AI system/bot learning specialists, as well as sensors, learning assessments, etc.
  • All learner data collected in the school, is stored in the learner’s LDV
  • If the learner enters any AI/AR/VR type learning environment, consent agreements are created instantly on the fly with the learner, school, school districts, learning specialists, etc.?
  • These specify how the learner will be identified, learning data use, storage, deletion, etc.
  • When the learner acquires learning credentials, these are digitally signed by the authoritative learning authority, and written to the learner’s SOLICT.
  • The SOLICT in turn pushes these out to the learner’s LSSI devices
  • The learner is now in control of their learning credentials
  • When the learner graduates, they’ll be able, with their consent, to offer use of their DLT, IEP and LDV to employers, post-secondary, etc.?This significantly reduces time and costs to train or help the learner learn
  • The learner continually leverages their DLT/IEP/LDV until their die i.e., it’s a lifelong learning system
  • IT’S TRANSFORMATIONAL OVER TIME, NOT OVERNIGHT

Tichaona Mhererwa

Founder & Principal Consultant @ Canvas dot Africa | Innovative Cloud, AI, and Software Solutions for Education, Finance, and Energy sectors.

2 个月

Quite a long read indeed. The following are some pertinent questions arising from this article, can we comprehensively have the same covered? 1. How can we ensure that AI agents designed for children are safe, age-appropriate, and transparent about their capabilities and limitations? 2. What guidelines or regulations should be in place to protect children's privacy and data when using AI agents? 3. How can parents and educators best guide children in navigating interactions with AI agents and maintain healthy boundaries? 4. What are the potential benefits and risks of using AI agents for educational purposes, such as tutoring or language learning? 5. How can we promote critical thinking and media literacy in children to help them understand the nature of AI agents and avoid being misled or manipulated?

回复
Harsh Mithaiwala

Web Developer @Altq | CS Grad at Concordia | Former Nokia Engineer | AI & Blockchain-Enthused

4 个月

Integrating AI agents into children's lives needs strong legal and privacy protections. Ensuring proper identity management and authorization is key for their safety and effective AI use. What safeguards can prevent unauthorized AI interactions with children?

要查看或添加评论,请登录

社区洞察

其他会员也浏览了