Sauvik Das

AI Is Taking Over Your Search Engine. Here's a Look Under the Hood ChatGPT, Gemini and company are going down the rabbit holes so you don't have to.

How does "AI search" differ from standard web search? Can you trust it?

I spoke to Jon Reed at CNET about this: www.cnet.com/tech/service...

The first (and only) rule of AI skillz club is that we don't talk about AI skillz, Mark.

see, that's why you're not a billionaire tech CEO. You've got good instincts

Hi all. I am delighted to announce our first Keynote speaker for the 2025 Summit on Responsible, Computing, AI, and Society: Dr. Rumman Chowdhury rcais.github.io

SOUPS 2025 Mentoring Program The Twenty-First Symposium on Usable Privacy and Security (SOUPS 2025), August 10–12, 2025, Seattle, WA, USA.

@sauvik.me and I are co-chairing the SOUPS mentoring!

If you are attending SOUPS, consider signing up to be a mentor or a mentee. It's a wonderful way to connect with others.
www.usenix.org/conference/s...

SOUPS 2025 Mentoring Program The Twenty-First Symposium on Usable Privacy and Security (SOUPS 2025), August 10–12, 2025, Seattle, WA, USA.

Are you going to be at the best conference, SOUPS (@soupsconference.bsky.social)?

If so, consider signing up to be a mentor or mentee for the mentoring program! It's a great way to meet other folks at the conference :) www.usenix.org/conference/s...

Who are all you people searching for my thumbprint?!

A major AI training data set contains millions of examples of personal data Personally identifiable information has been found in DataComp CommonPool, one of the largest open-source data sets used to train image generation models.

One of the largest text-image datasets is full of PII, including credit card numbers and birth certificates. Excellent writeup by @eileenguo.bsky.social www.technologyreview.com/2025/07/18/1... Read our full audit and legal analysis arxiv.org/pdf/2506.17185

You see, @frank.computer, the problem is not wanting to welcome rapists is too one-sided.

We need to hear, too, from those who *do* want to welcome rapists. It's only fair

I've been receiving a new kind of email lately: people who GPT has convinced to email professors in order to get their "deeply insightful" conversations w/ GPT published.

It's a weird confluence of AI sycophancy + human loneliness and delusions of grandeur.

Maybe he was going for a Dreyfussian Man?

Usability is like security: if you don't design around it from the very start, when you try to patch it on, unless you get lucky, you'll discover all kinds of architectural commitments that make it really hard, and you'll always play catch-up.

It's even more fun when you are trying to retrofit usability into your retrofitted security!

As of today, I am officially Associate Professor* at @hcii.cmu.edu. Surreal — I was out of my depth when I started. Couldn't have done it without my incredible students — the true brains of the operation — and the support of my mentors. Thank you all!

* not yet tenured, because CMU

Navigating Security and Privacy Threats in Homeless Service Provision People experiencing homelessness interact with service providers to access essential services. As clients, homeless individuals are expected to reveal sensitive information about themselves to service providers, while personal security and privacy (S&P) preferences fall by the wayside. Simultaneously, providers take on S&P-adjacent responsibilities: helping clients fill out applications, safekeeping clients’ personal documents, monitoring clients’ online safety, undergoing workplace S&P training, etc. We created five storyboards to represent S&P challenges that clients can face when they interact with providers. In interviews with homeless individuals and service providers in the Northeastern United States, we use these storyboards to explore the S&P challenges in client-provider relationships within homeless services. We find a set of mismatches in S&P priorities between clients and providers, leading to mistrust between the two parties. We provide design recommendations, envisioned by both parties, for providers to bridge these mismatches.

new paper incoming at USENIX Security 2025 📬

"Navigating Security and Privacy Threats in Homeless Service Provision"

really grateful to have worked with @hongshen.bsky.social, Fei Fang, and a team of amazing undergrads and community collaborators on this project

yuxiwu.com/pubs/sec25_h...

Quite a notification from LinkedIn...@mmazurek.bsky.social

Hyped for HAIPS 2025! 👏
haips.com

#HAIPS2025 #HumanCenteredAI #Privacy #Security

📢 We now have two amazing keynote speakers lined up for #HAIPS2025 (the 1st Workshop on Human-Centered AI, Privacy, and Security)! @jas0nh0ng.bsky.social
and @pgk.bsky.social

Submit your:
- Original papers
- Encore papers
- SoK papers
- Vision papers

by June 20th!

💻http://haips.com

🚨 New Paper! 🚨

“Groups that communicate diverse information make wiser decisions.” Right? Not always—for LLMs, just like humans.

We bring this scrutiny to AI by introducing the Hidden Profile paradigm to assess how multi-agent LLMs actually reason together.

arxiv.org/abs/2505.115...
[1/n]

I'm sending a harmonica to your house as we bleet

😬In sims, 1/100 Asian-coded agents will join a protest if told they shouldn't. All 100 Black-coded agents will join anyway.

New #Facct2025 paper led by @yuxuanli1225.bsky.social
Actions Speak Louder than Words: Agent Decisions Reveal Implicit Biases in Language Models

sauvikdas.com/papers/64/se...

You People Made Me Give Up My Peanut Farm Before I Got To Be President

From The Archives: You People Made Me Give Up My Peanut Farm Before I Got To Be President theonion.com/you-peo...

Announcing the 1st Workshop on Human-Centered AI, Privacy, and Security (#HAIPS2025) at #CCS2025

Send us your tired, your poor, Your huddled masses yearning to breathe free, and papers you just had rejected from #usesec25 or #ICML2025

📢 The CFP is out!

🌊 haips.com

#AcademicSky

This is just the yin and yang of working on emerging tech. Critical perspectives often get attention only after an initial hype cycle because systemic impacts become more obvious post-deployment. There isn't much (known) critical scholarship on, e.g., quantum computing for this reason

AI is a meta-cycle (in that it spans many hype cycles), but it's helpful to contextualize it within broader trends. When I started my Ph.D., the "emerging" tech hype centered on social media and smartphones. At the beginning, it was mostly positive. But critical perspectives gradually won attention.

Purpose Mode helps social media users stay focused on their desired tasks A CMU research team has developed Purpose Mode — a browser extension that allows users to “toggle off” Attention Capture Damaging Patterns (ACDPs) while using social media platforms.

@cylab.bsky.social wrote a great article on @hankhplee.bsky.social's #chi2025 paper on Purpose Mode.

He presented it today and did a great job!

PurposeMode reduces distraction on social media from 28% -> 7%. You can use it, too! It's open source :)

cylab.cmu.edu/news/2025/04...

Illustration of a city with a ferris wheel, and there is a tiny boat in the water that says "CHI". Text reads CHI 2025, April 26-May 1, 2025 in Yokohama, Japan.

Yokohama (and virtual attendance), here we come! ✈️

We're excited to share that @cmu.edu authors contributed to 50+ papers (with 12 award-winning!) at the #CHI2025 Conference.

Check out this overview of our involvement here
➡️ hcii.cmu.edu/news/cmu-chi...

Stay tuned, more CHI highlights to come.

Wednesday: Farida Elsehin (MS grad from CMU, soon to be Ph.D. student at UNCC) will be presenting her work on a mixed-methods study to unpack what cryptocurrency users understand about seed phrases.

Headline: Most harbor significant misconceptions.

bsky.app/profile/sauv...

Wednesday: @hankhplee.bsky.social
will be presenting our work on developing "Purpose Mode", a browser extension that reduces distraction on social media from 28% to 7%.

bsky.app/profile/sauv...

2/3

En route to #chi2025!

1) Sunday: I will present a keynote talk at the computational UI workshop — the first public discussion of what we're doing at fuguUX, the company I'm co-founding with @jas0nh0ng.bsky.social
.

"Towards Smart, Automated UI Assessments at Scale"

1/3