JavaScript is not available.

Oct 25, 2022

So far, I've read 707 articles for my Ph.D. (#privacy, #dataprotection law, #fairness at

@TAU_LAW

). Below are some of the articles that have most influenced my research: (and they are free) *The list is not in order of preference **For the book thread, see the last Tweet

204

803

Sign up for The Privacy Whisperer and receive my weekly analysis of topics like this: lnkd.in/dizAKhip. If you like #privacy threads, follow me here. 12/12

Privacy insights & tips every week. Join our growing community of 58,000+ privacy supporters. Click to read The Privacy Whisperer, by Luiza Jarovsky, a Substack publication with thousands of readers.

The Privacy Whisperer

My opinion: legislation should be much more stringent in protecting humans: autonomy, privacy, transparency, and fairness. This is the wild west of algorithms, and it will get worse. 11/12

What's the common denominator - or how to get to 100 million users quicker? Concerns about human values, limitations, biases, and the risk of causing harm to individuals or societies are not the focus. Autonomy and broad privacy concerns are not priorities. 10/12

and possibly generating reputational harm. Its potential for bias and misinformation is immense, as the tool is trained to be persuasive and imitate human discourse, not to distinguish fact from lies. It can also infringe copyright and protected information. 9/12

If you have ever posted anything in English on the internet, it is possible that your data was used to train it. It can be "attacked" and made to disclose personal data about people, sometimes true (scrapped from the internet), sometimes fake, per association, 8/12

As I discussed in my newsletter, "ChatGPT And Large Language Models Are A Privacy Ticking Bomb", it's based on large language models and trained using publicly available data - containing massive amounts of personal data. theprivacywhisperer.com/p/chatgpt-and- 7/12

ChatGPT And Large Language Models Are A Privacy Ticking Bomb

Hi everyone, I hope you are doing well. In today's newsletter, I discuss some of the emerging privacy issues associated with tools based on large language models - AI systems such as ChatGPT....

ChatGPT - or Chat Generative Pre-Trained Transformer - the tool everyone is talking about, is a new generation of tech. It can answer questions, create songs, articles, summaries, code, and even pass an MBA exam. 6/12

It also uses dark patterns in privacy to make people share more data, as I frequently discuss here. twitter.com/LuizaJarovsky/ 5/12

Quote Tweet

Feb 2

TikTok and #darkpatterns in #privacy TikTok is always educational when teaching us about dark patterns in the context of privacy. Let's take a look at this popup (a cluster of dark patterns):

(1/13)

Show this thread

It is built to trigger compulsive use, especially in more impressionable audiences such as teenagers. Its algorithms leave no space for autonomy, reflexive choice, or critical thinking: everything is fast and frenetic and must impress the user in 3 seconds. 4/12

As I wrote in my newsletter article "I Was On TikTok For 30 Days: It Is Manipulative, Addictive, And Harmful To Privacy", it uses manipulative user experience (UX) design to keep users glued to it. theprivacywhisperer.com/p/i-was-on-tik 3/12

I Was On TikTok For 30 Days: It Is Manipulative, Addictive, And Harmful To Privacy

Whether you like it or not, TikTok’s numbers are impressive. It has over 1 billion monthly active users. It was downloaded 3 billion times worldwide, and it is so far the most downloaded app this...

As you can see above, it's getting "easier" to reach 100 million users. It took #TikTok 9 months; #ChatGPT only 2 months. What's changing? Besides globalization and the internet, algorithms are getting more powerful and tailored to impress, manipulate, and overcome humans. 2/12

🔥 TikTok, ChatGPT, and #Privacy A thread 🧵1/12

Subscribe to The Privacy Whisperer to receive #privacy insights and tips every week in your inbox: theprivacywhisperer.com (13/13)

Privacy insights & tips every week. Join our growing community of 58,000+ privacy supporters. Click to read The Privacy Whisperer, by Luiza Jarovsky, a Substack publication with thousands of readers.

The Privacy Whisperer

Third, my newsletter article "Dark Patterns (Deceptive Design) in Data Protection" theprivacywhisperer.com/p/dark-pattern (12/13)

Dark Patterns (Deceptive Design) in Data Protection

Learn how to identify and stop them from dictating your online choices

Second, my newsletter article: "I Was On TikTok For 30 Days: It Is Manipulative, Addictive, And Harmful To Privacy" theprivacywhisperer.com/p/i-was-on-tik (11/13)

I Was On TikTok For 30 Days: It Is Manipulative, Addictive, And Harmful To Privacy

If you want to learn more about dark patterns in privacy, here are some resources. First, my academic paper: "Dark Patterns in Personal Data Collection: Definition, Taxonomy and Lawfulness" - papers.ssrn.com/sol3/papers.cf (10/13)

6- Does email access include all my email content or only the email contact list? It's not clear. (9/13)

5. Why are the buttons asymmetric: "OK" in capital letters and bold and "Don't allow" not capitalized and not in bold. Why not "allow" and "don't allow" (nothing in bold) to make it clear that there are permissions being asked and that I have a choice? (8/13)

4- Why are 2 different data processing purposes (experience improvement and personalized ads) under one "OK"? Shouldn't I have separate choices over my data? (7/13)

3. Why the 2 calls to action are under one "OK"? Shouldn't I have separate choices over my data? (6/13)

2. Why are personalized ads described as experience improvement? Is every additional access now disguised as experience improvement? (5/13)

6 questions: 1. What is the connection between giving access to Facebook friends list and giving access to email? Why ask these two permissions in the same privacy notice if these are two totally different contexts and types of data? (4/13)

Explanations/assumptions - It says that doing the above will lead to experience improvement; - It says that connecting me with friends (from email and Facebook) is a form of experience improvement; - It says that ad personalization is a form of experience improvement (3/13)

In the same popup, there are 2 calls to action and 3 different explanations/assumptions: Calls to action: - It asks me to give access to my Facebook friends list - It asks me to give access to my email (2/13)

🔥TikTok and #darkpatterns in #privacy TikTok is always educational when teaching us about dark patterns in the context of privacy. Let's take a look at this popup (a cluster of dark patterns): 🧵 (1/13)

Read my full newsletter on the topic (subscribe using the same link): "ChatGPT And Large Language Models Are A Privacy Ticking Bomb" theprivacywhisperer.com/p/chatgpt-and- It would be great to hear your views! 22/22

ChatGPT And Large Language Models Are A Privacy Ticking Bomb

Hi everyone, I hope you are doing well. In today's newsletter, I discuss some of the emerging privacy issues associated with tools based on large language models - AI systems such as ChatGPT....

These models are constantly being trained, and the more they are used, the more "advanced" will be their next version - including in terms of potential privacy harms - if we do not do anything about it. 21/22

I never gave my data to LLM-based chatbots; I never consented that my data was going to be scrapped from the internet, condensed in a context-less way, sometimes fantasized or distorted, then potentially output by a chatbot when anyone around the world gives it a prompt. 20/22

I do not even need to go further and analyze specific articles to point out that it is unclear to me how LLM-based tools can be GDPR compliant. 19/22

If we remember GDPR's article 5, it brings principles relating to the processing of personal data. Among them are purpose limitation, data minimization, and accuracy. 18/22

Aren't we going back in time and neglecting privacy rights that it took us so long to achieve? 17/22

If people are going to resort to these tools to obtain information, does it make sense that it is going to output fake information, including fantasized personal information - causing privacy and reputational harm? 16/22

So in a similar sense, if someone is constantly writing about crime or murder news, this person's name might be associated with criminal offenses, causing reputational harm. 15/22

He has tweeted about his love of dogs in the past, but he doesn't seem to have any of his own.'(Incorrect.)." In this case, some of the information was correct, but some was pure fantasy, probably some form of association. 14/22

"told me Mat has a wife and two young daughters (correct, apart from the names), and lives in San Francisco (correct). It also told me it wasn’t sure if Mat has a dog: '[From] what we can see on social media, it doesn't appear that Mat Honan has any pets. 13/22

In her article for MIT Technology Review,

@Melissahei

describes that after prompting GPT-3 with questions about her boss, the tool technologyreview.com/2022/08/31/105 11/22

technologyreview.com

What does GPT-3 “know” about me?

Large language models are trained on troves of personal data hoovered from the internet. So I wanted to know: What does it have on me?

You would not want that your inferred address or information about a life milestone that you posted on a social network would be shared with others by a chatbot, e.g., when prompted to create a marketing campaign or to offer customer support. 10/22

Even if the information was "publicly available," this is a breach of contextual integrity, a concept popularized by

@HNissenbaum

and central to privacy. When you post something personal online, it is necessarily connected to the context you chose to post. 9/22

Through their "attack" on GPT-2, they were able to extract "personally identifiable information (names, phone numbers, and email addresses), IRC conversations, code, and 128-bit UUIDs." arxiv.org/pdf/2012.07805 8/22