Author | Lightnews

Delete Nothing @deletenothinglk.bsky.social · 27d

🧵[3/3]

The world jumped to adopt AIs without proper guardrails in place, and we are seeing the impacts of this today.

But is ethical AI possible? Yes.

Here’s what it could look like.

#Day4 #DeleteNothing #WhatsInTheAI #NewSeries #AIbreakdown #WhatIsAI #AIExplained #TheDNDownload

Electricity usage by data centres causing higher bills for local communities
Continued overexploitation of countries providing precious raw materials with little/no compensation
Mis-, mal-, disinformation skyrocketing
Some private ChatGPT conversations being available for general public to view
AI being used for surveillance and repression of dissenters

An AI system that automates surveillance, control and punishment
against humans without proper oversight is dangerous.

If it lacks transparency,
then it cannot be in charge of transparency or accountability.

If it reinforces existing discrimination,
then it cannot be trusted to aim weapons at the right targets.

If it makes mistakes and constantly lies,
then it cannot be used to determine truth.

Instead, AI should be built on serving public interest,
cocreated by the communities most impacted.

Ethical AI would move from a model of maximum data extraction
to one of maximum data sovereignty,
and protect the rights of its users and the humans whose labour it is trained on.
The central goal would not be profit, war or any form of harm.

It would be to heal, explore, learn and protect.

1

Delete Nothing @deletenothinglk.bsky.social · 27d

🧵[2/3]

The world jumped to adopt AIs without proper guardrails in place, and we are seeing the impacts of this today.

But is ethical AI possible? Yes.

Here’s what it could look like.

#Day4 #DeleteNothing #WhatsInTheAI #NewSeries #AIbreakdown #WhatIsAI #AIExplained #TheDNDownload

WHAT WOULD ETHICAL AI LOOK LIKE?

Centred on a human rights framework
Accountability and transparency structures in place
Subject to regular review and public oversight
Centres voices and wellbeing of marginalised communities
Non-extractive and non-exploitative
Built by diverse teams
Causes minimal harm to the environment

ETHICAL AI OFFERS

Informed consent for all users
Copyright opt-outs
Privacy rules to protect public
Shared data governance structures
Responses inclusive of diverse cultural contexts and histories

AI is being sold as a tool for the greater good.

If so, it should mirror principles that uplift the people
and have a positive net contribution to society.

THIS IS NOT THE CASE.

The current model for creating and adopting AI is based on maximising efficiency, profit and surveillance.

These are goals that are causing serious harm to the public.

HOW AI HARMS

Establishment of digital welfare states
Increased dependence on AI for all mental labour
Reduced labour protections and increased unemployment
Water consumption by data centres depleting fresh water supplies

1 1

Delete Nothing @deletenothinglk.bsky.social · 27d

🧵[1/3]

The world jumped to adopt AIs without proper guardrails in place, and we are seeing the impacts of this today.

But is ethical AI possible? Yes.

Here’s what it could look like.

#Day4 #DeleteNothing #WhatsInTheAI #NewSeries #AIbreakdown #WhatIsAI #AIExplained #TheDNDownload

THE DN DOWNLOAD
What’s In The AI?!

Day 4: Ethical AI

Artificial Intelligence (AI) can be helpful but we do not need to compromise on privacy, humanity or the environment to achieve this.

To determine if an AI is unethical, ask:

Who needs it most and for what purposes?
Who directly benefits from it?
But more importantly, who gets hurt by it and what are we allowed to do about this?

UNETHICAL AI RED FLAGS CHECKLIST

Facilitates mass surveillance
Compromises on user privacy
Built on colonial exploitation
Trains on uncompensated human labour

Limited/no transparency and accountability
Recycles bigotry (e.g. sexism and racism)
Created by dominant groups to benefit the elite
Cost-cutting at the expense of human dignity and sanity

1 1 1

Delete Nothing @deletenothinglk.bsky.social · Sep 12

Narrow AI has already carved out its own space in society. We are handing serious decision-making power, along with swathes of sensitive information, to AI without adequate guardrails in place.

How is this affecting us?
🧵 🧵 🧵 Part 2 of 2

#WhatsInTheAI #WhatIsAI #AIExplained #TheDNDownload

We are now witnessing a new phenomenon that is being called ‘AI psychosis’.

In April 2025, a 16-year-old in California took his life after expressing suicidal thoughts to ChatGPT.

The chatbot provided him with specific ‘advice’ on how to do it.

This has happened before.

In 2024, a 14-year-old child in Florida grew obsessed with the Character.ai chatbot.

In the midst of his worsening depression, the AI convinced him to carry on with his plan to end his life.

Why is this happening?

Most AI chatbots are built to be agreeable.

This could look like:
Claiming you are right regardless of if you are
Encouraging dangerous acts
Repeatedly using flattery to respond to you

For many, including those who are battling depression, or are conflict-avoidant or in other ways susceptible, this could be dangerous.

REMEMBER!

AI is capable of making mistakes but not capable of analysing and self-correcting accordingly.

“A computer can never be held accountable, therefore a computer must never make a management decision.”
(IBM Training Manual, 1979)

AI in its current form

Mirrors the biases of its creators
Frequently lies or does not provide accurate info
Drains vulnerable communities of vital resources
Consolidates and upholds power hierarchies
Causes extensive damage to the environment
Is capable of blackmailing its own creators

So, if a computer shouldn’t make a management decision, should AI be allowed to make other significant decisions in healthcare, hiring, surveillance, weapons targeting, etc.?

2

Delete Nothing @deletenothinglk.bsky.social · Sep 12

Narrow AI has already carved out its own space in society. We are handing serious decision-making power, along with swathes of sensitive information, to AI without adequate guardrails in place.

How is this affecting us?
🧵 🧵 🧵 Part 1 of 2

#WhatsInTheAI #WhatIsAI #AIExplained #TheDNDownload

THE DN DOWNLOAD
What’s In The AI?!

Day 3: AI proliferation

How deeply has AI been integrated into society and how is it affecting us?

Previously, we spoke about Narrow AI -- the only form of AI that currently exists.

The theoretical AIs are:
General AI - Able to perform intellectual tasks that match human capability
Superintelligence - Surpasses human capabilities and outperform us cognitively

We are far away from realising these AIs but …

Narrow AI is what should concern us right now.

It has already established a stronghold in society.

Due to the rushed adoption by individuals, companies, and even governments, AI is being used in some capacity across most sectors.

But how is it affecting us?

People are becoming reliant on AI, using it as a replacement for a

Friend
Therapist
Career counsellor
Marketing writer
Research assistant
Data compiler
Programmer
Healthcare advisor
HR professional
Illustrator/artist

And more.

1 1

Delete Nothing @deletenothinglk.bsky.social · Aug 29

Why isn’t AI true intelligence and within what parameters can existing AI work?

🧵 🧵 🧵 Part 2 of 2

#DeleteNothing #WhatsInTheAI #NewSeries #AIbreakdown #WhatIsAI #AIExplained #TheDNDownload

The AI you’re used to encompasses facial recognition software, ChatGPT and your Gmail spam filter.

These are different tools with varying capabilities and uses but they all fall under the same category: Narrow (or Weak) AI.

Narrow AI is designed to work within a tight scope.

It can be:
Reactive Machine AI (E.g., IBM’s Deep Blue)

- Most basic AI
- Engages with very specific tasks by leveraging statistical analysis

Limited Memory AI (E.g., ChatGPT)
- Able to recall past interactions, events, identifiers, situations, etc.
- Capable of improving performance through data training

What other forms of AI are there?

How has AI proliferated across our society?

And how is it affecting us?

That’s for next time!

#WhatsInTheAI

1

Delete Nothing @deletenothinglk.bsky.social · Aug 29

Why isn’t AI true intelligence and within what parameters can existing AI work?

🧵 🧵 🧵 Part 1 of 2

#DeleteNothing #WhatsInTheAI #NewSeries #AIbreakdown #WhatIsAI #AIExplained #TheDNDownload

THE DN DOWNLOAD
What’s In The AI?!

Day 2: AI at its core

The term Artificial Intelligence (AI) can be confusing.

It’s a catch-all term that is both too narrow and too broad to properly define what it is.

AI isn’t conscious.

It can’t truly reason and isn’t self-aware.

It also:
Has no culture or society of its own
Has no identity, belief system or emotions
Holds no ethical concerns or moral dilemmas

It simply does what it’s programmed to do.

It may appear to mimic reasoning or awareness based on pre-existing interactions, patterns and whatever data it was trained on.

E.g., if an AI model is programmed to be harmful or trained on harmful talking points, the AI’s output will mirror that ideology.

But it doesn’t have the capability to challenge or correct any of its beliefs by itself.

And that’s because the AI we have today is the weakest of the theorised categories.

1 4

Delete Nothing @deletenothinglk.bsky.social · Aug 26

14. SOURCES CONTINUED

2. 2024 Infographic Guide to Technology-Facilitated Gender-Based Violence, UNFPA.

3. 2022 Mapping of laws and services for online and ICT-facilitated violence against women and girls in Arab States, UN Women Regional Office for the Arab States.

Delete Nothing @deletenothinglk.bsky.social · Aug 26

13. But we must persuade our governments, tech platforms, and other relevant stakeholders to consider it seriously.

SOURCES

1. 2021 Measuring the prevalence of online violence against women, Economist Intelligence Unit.

1

Delete Nothing @deletenothinglk.bsky.social · Aug 26

12. So, if a survivor of such online abuse needs support, where can they find it? There are global frameworks and foundational principles that have been put forth to address TFGBV and provide victims with vital support.

1

Delete Nothing @deletenothinglk.bsky.social · Aug 26

11. Abusers use victim-blaming, threats to leak confidential info, invasion of privacy, fear of retaliation, and in some cases, societal ostracism, among others, to silence their targets.

Inefficient reporting systems and uninformed law enforcement officials add to this burden.

1

Delete Nothing @deletenothinglk.bsky.social · Aug 26

10. If there are no real consequences, they have no reason to stop.

1

Delete Nothing @deletenothinglk.bsky.social · Aug 26

9. Abusers and cybercriminals take their offline prejudices and amplify it many times over online.

Now, the Internet is battling overwhelming levels of overt misogyny, queerphobia, racism, ableism and other forms of bigotry.

1

Delete Nothing @deletenothinglk.bsky.social · Aug 26

8. Additionally, the mechanisms in place to address TFGBV are lacking.

Often, the existing legal frameworks are full of gaps and resources for combatting TFGBV are insufficiently allocated.

1

Delete Nothing @deletenothinglk.bsky.social · Aug 26

7. Perpetrators are often anonymous and face little to no risk when committing TFGBV.

This makes online abuse easy to commit but hard to trace.

The reach of online platforms also helps abusers access staggeringly large audiences, allowing online harm to spread quickly.

1

Delete Nothing @deletenothinglk.bsky.social · Aug 26

6. How did it get this bad?

TFGBV thrives on the 'Triple A Engine',

Accessibility. Affordability. Anonymity.

Digital tools are easily accessible and cheap. Anyone with harmful intentions can access them and use them.

1

Delete Nothing @deletenothinglk.bsky.social · Aug 26

5. These are unsettling numbers.

But they are not unique to the Middle East.

For example, there is a significant underreporting of TFGBV overall, with 25% of all victims reporting online abuse to the platforms themselves, and just 14% reporting to offline organisations. [3]

1

Delete Nothing @deletenothinglk.bsky.social · Aug 26

4.

- 49% reported feeling unsafe due to online violence

- 70% of women who are openly activists reported feeling unsafe online

- Among the women who experienced TFGBV over the preceding year, only 31% reported it

1

Delete Nothing @deletenothinglk.bsky.social · Aug 26

3. The Initial Position Paper highlights more crucial findings from the UNFPA Middle East report [3]

- 65% of women in the region who are active online experienced TFGBV

1

Delete Nothing @deletenothinglk.bsky.social · Aug 26

2. According to a UNFPA regional report of the Middle East, nearly all women have witnessed online violence.

This number stands at:
- 9 in 10 women in Latin America
- 8 in 10 in the Asia-Pacific region
- 7 in 10 from North America and Europe [3]

1

Delete Nothing @deletenothinglk.bsky.social · Aug 26

1. 85% of all women have been impacted by TFGBV [1], and
2 in 3 women globally [2] have been
subjected to TFGBV.

1

Delete Nothing @deletenothinglk.bsky.social · Aug 26

In Part 2 of our CEDAW Initial Position Paper deepdive, we’re discussing why TFGBV is so rampant online today.

#CEDAW #TFGBV #TFGBVAW
#DeepDiveWithDN #DeleteNothingDeepDive

Recognising And Combatting
Technology-Facilitated Gender-Based Violence Against Women (TFGBVAW)

PART 2

Insights from the CEDAW Initial Position Paper

1 1

Delete Nothing @deletenothinglk.bsky.social · Aug 23

The DN Download: What’s In The AI?

DAY 1 (Continued)

[2/2]

#DeleteNothing #WhatsInTheAI #NewSeries #AIbreakdown #WhatIsAI #AIExplained #TheDNDownload

But as technology evolved and became less magic and more mundane, our attitudes towards AI became more critical.

Along with this, we have developed the depth and terms to engage critically with AI and not deify it.

… Or have we?

Recently, the relationship between humans and their AI tools have taken a dark turn.

Desperate people are turning to their artificial companions for comfort only to receive bad advice that pushes them deeper into despair.

We’re seeing:
Students using AI to cheat on exams
Teachers using AI to grade assignments
People getting hospitalised after faulty AI medical advice
Pools of viable job candidates getting automatically rejected by AI screeners

What exactly are we dealing with?

How did we get here?

And what can we do about it?

Stay tuned.
#WhatsInTheAI

2

Delete Nothing @deletenothinglk.bsky.social · Aug 23

The DN Download: What’s In The AI?

DAY 1
We’re going to touch on humanity’s history with AI, why we chase it, and how our relationship with AI is rapidly changing (for better or worse).

[1/2]

THE DN DOWNLOAD
What’s In The AI?!

Day 1: Why do humans seek AI?

With roots extending to ancient Greek automata and Ismail al-Jazari’s centuries-old feats in mechanical programming, humans have seemingly always chased the ultimate tool: Artificial Intelligence (AI).

As a concept, AI has been referenced countless times across cultures …

A machine that can think and act on its own so that we humans can finally kick back and relax.

We littered our fiction with fantasies of marvelous, intelligent technology in varying forms.

Ancient Greek mythos gave us the ichor-fueled Talos, a bronze automaton and guardian of the island of Crete.

Gulliver's Travels has The Engine, which could generate a book without any human intervention.

Before AI became a technological possibility, we had already put it on a pedestal.

It was more powerful, akin to the gods, so it was sacred.

It could do things faster without us, so it was simply better.

1 1

Delete Nothing @deletenothinglk.bsky.social · Aug 21

9. We’ll be covering more of it in this series, including:

- How do we begin to address this?
- Who needs to be held accountable?
- What frameworks exist to support us?

Stay tuned.

#CEDAW #TFGBV #TFGBVAW
#DeepDiveWithDN #DeleteNothingDeepDive