AI Insights
Apple Workshop on Privacy-Preserving Machine Learning 2025
Apple believes that privacy is a fundamental human right. As AI experiences become increasingly personal and a part of people’s daily lives, it’s important that novel privacy-preserving techniques are created in parallel to advancing AI capabilities.
Apple’s fundamental research has consistently pushed the state-of-the-art in using differential privacy with machine learning, and earlier this year, we hosted the Workshop on Privacy-Preserving Machine Learning (PPML). This two-day hybrid event brought together Apple and members of the broader research community to discuss the state of the art in PPML, focusing on four key areas: Private Learning and Statistics, Attacks and Security, Differential Privacy Foundations, and Foundation Models and Privacy.
The presentations and discussions of these topics explored the intersection of privacy, security, and the rapidly evolving landscape of artificial intelligence. Workshop participants discussed the theoretical underpinnings and practical challenges of building AI systems that protect privacy. By addressing privacy and security concerns from both theoretical and practical perspectives, we aim to foster innovation while safeguarding user privacy.
In this post, we share recordings of selected talks and a recap of the publications discussed at the workshop.
Apple Workshop on Privacy-Preserving Machine Learning 2025 Videos
Published Work Presented at the Workshop
AirGapAgent: Protecting Privacy-Conscious Conversational Agents by Eugene Bagdasarian (Google Research), Peter Kairouz (Google Research), Ren Yi (Google Research), Marco Gruteser (Google Research), Sahra Ghalebikesabi (Google DeepMind), Sewoong Oh (Google Research), Borja Balle (Google DeepMind), and Daniel Ramage (Google Research)
A Generalized Binary Tree Mechanism for Differentially Private Approximation of All-Pair Distances by Michael Dinitz (Johns Hopkins University), Chenglin Fan (Seoul National University), Jingcheng Liu (Nanjing University), Jalaj Upadhyay (Rutgers University), and Zongrui Zou (Nanjing University)
Differentially Private Synthetic Data via Foundation Model APIs 1: Images by Zinan Lin (Microsoft Research), Sivakanth Gopi (Microsoft Research), Janardhan Kulkarni (Microsoft Research), Harsha Nori (Microsoft Research), and Sergey Yekhanin (Microsoft Research)
Differentially Private Synthetic Data via Foundation Model APIs 2: Text by Chulin Xie (University of Illinois Urbana-Champaign), Zinan Lin (Microsoft Research), Arturs Backurs (Microsoft Research), Sivakanth Gopi (Microsoft Research), Da Yu (Sun Yat-sen University), Huseyin Inan (Microsoft Research), Harsha Nori (Microsoft Research), Haotian Jiang (Microsoft Research), Huishuai Zhang (Microsoft Research), Yin Tat Lee (Microsoft Research), Bo Li (University of Illinois Urbana-Champaign, University of Chicago), and Sergey Yekhanin (Microsoft Research)
Efficient and Near-Optimal Noise Generation for Streaming Differential Privacy by Krishnamurthy (Dj) Dvijotham (Google DeepMind), H. Brendan McMahan (Google Research), Krishna Pillutla (ITT Madras), Thomas Steinke (Google DeepMind), and Abhradeep Thakurta (Google DeepMind)
Elephants Do Not Forget: Differential Privacy with State Continuity for Privacy Budget by Jiankai Jin (The University of Melbourne), Chitchanok Chuengsatiansup (The University of Melbourne), Toby Murray (The University of Melbourne), Benjamin I. P. Rubinstein (The University of Melbourne), Yuval Yarom (Ruhr University Bochum), and Olga Ohrimenko (The University of Melbourne)
Improved Differentially Private Continual Observation Using Group Algebra by Monika Henzinger (Institute of Science and Technology (ISTA) Austria) and Jalaj Upadhyay (Rutgers University)
Instance-Optimal Private Density Estimation in the Wasserstein Distance by Vitaly Feldman, Audra McMillan, Satchit Sivakumar (Boston University), and Kunal Talwar
Leveraging Model Guidance to Extract Training Data from Personalized Diffusion Models by Xiaoyu Wu (Carnegie Mellon University), Jiaru Zhang (Purdue University), and Steven Wu (Carnegie Mellon University)
Local Pan-privacy for Federated Analytics by Vitaly Feldman, Audra McMillan, Guy N. Rothblum, and Kunal Talwar
Nearly Tight Black-Box Auditing of Differentially Private Machine Learning by Meenatchi Sundaram Muthu Selva Annamalai (University College London) and Emiliano De Cristofaro (University of California, Riverside)
On the Price of Differential Privacy for Hierarchical Clustering by Chengyuan Deng (Rutgers University), Jie Gao (Rutgers University), Jalaj Upadhyay (Rutgers University), Chen Wang (Texas A&M University), and Samson Zhou (Texas A&M University)
Operationalizing Contextual Integrity in Privacy-Conscious Assistants by Sahra Ghalebikesabi (Google DeepMind), Eugene Bagdasaryan (Google Research), Ren Yi (Google Research), Itay Yona (Google DeepMind), Ilia Shumailov (Google DeepMind), Aneesh Pappu (Google DeepMind), Chongyang Shi (Google DeepMind), Laura Weidinger (Google DeepMind), Robert Stanforth (Google DeepMind), Leonard Berrada (Google DeepMind), Pushmeet Kohli (Google DeepMind), Po-Sen Huang (Google DeepMind), and Borja Balle (Google DeepMind)
PREAMBLE: Private and Efficient Aggregation via Block Sparse Vectors by Hilal Asi, Vitaly Feldman, Hannah Keller (Aarhus Univiersity; work done while at Apple), Guy N. Rothblum, Kunal Talwar
Privacy amplification by random allocation by Vitaly Feldman (Apple) and Moshe Shenfeld (The Hebrew University of Jerusalem)
Privacy of Noisy Stochastic Gradient Descent: More Iterations without More Privacy Loss by Jason Altschuler (MIT) and Kunal Talwar
Privately Estimating a Single Parameter by John Duchi (Stanford University), Hilal Ali, and Kunal Talwar
Scalable Private Search with Wally by Hilal Asi, Fabian Boemer, Nicholas Genise, Muhammad Haris Mughees, Tabitha Ogilvie, Rehan Rishi, Guy N. Rothblum, Kunal Talwar, Karl Tarbe, Ruiyu Zhu, and Marco Zuliani
Shifted Composition I: Harnack and Reverse Transport Inequalities by Jason Altschuler (University of Pennsylvania) and Sinho Chewi (IAS)
Shifted Interpolation for Differential Privacy by Jinho Bok (University of Pennsylvania), Weijie Su (University of Pennsylvania), and Jason Altschuler (University of Pennsylvania)
Tractable Agreement Protocols by Natalie Collina (University of Pennsylvania), Surbhi Goel (University of Pennsylvania), Varun Gupta (University of Pennsylvania), and Aaron Roth (University of Pennsylvania)
Tukey Depth Mechanisms for Practical Private Mean Estimation by Gavin Brown (University of Washington) and Lydia Zakynthinou (University of California, Berkeley)
User Inference Attacks on Large Language Models by Nikhil Kandpal (University of Toronto & Vector Institute), Krishna Pillutla (Google), Alina Oprea (Google, Northeastern University), Peter Kairouz (Google), Christopher A. Choquette-Choo (Google), and Zheng Xu (Google)
Universally Instance-Optimal Mechanisms for Private Statistical Estimation by Hilal Asi, John C. Duchi (Stanford University), Saminul Haque (Stanford University), Zewei Li (Northwestern University), and Feng Ruan (Northwestern University)
“What do you want from theory alone?” Experimenting with Tight Auditing of Differentially Private Synthetic Data Generation by Meenatchi Sundaram Muthu Selva Annamalai (University College London), Georgi Ganev (University College London, Hazy), and Emiliano De Cristofaro (University of California, Riverside)
Acknowledgments
Many people contributed to this workshop including Hilal Asi, Anthony Chivetta, Vitaly Feldman, Haris Mughees, Martin Pelikan, Rehan Rishi, Guy Rothblum, and Kunal Talwar.
AI Insights
2 Artificial Intelligence (AI) Stocks That Could Become $1 Trillion Giants

These AI growth stocks may still be undervalued on Wall Street.
There are 10 companies with a market cap over $1 trillion right now, and all of these except one are involved in artificial intelligence (AI). This technology will drive a substantial amount of economic growth in the 21st century, providing investors the chance to earn substantial gains from the right stocks.
Some companies that are well positioned to play a key role in shaping the economy with AI are still valued at less than $1 trillion. Although their share prices could be volatile in the near term, the following two companies could be worth a lot more down the road they are today.
Image source: Getty Images.
1. Palantir Technologies
More than 800 companies have chosen Palantir Technologies (PLTR 4.14%) to transform their business operations with AI. Businesses can upload data on Palantir’s platforms, and it basically shows them how to be more efficient, grow their revenue, and become more profitable. It is working magic for businesses and the U.S. military, which trusts Palantir to keep top-secret information secure about the U.S. and its allies. Despite its already high market cap of $400 billion, Palantir’s unique value proposition and stellar profitability has all the makings of a $1 trillion business.
Palantir is not just slapping a large language model on a company’s data to make it easy to search information. It pulls together data from different sources within a company, which creates a framework for understanding how the company operates. Palantir is essentially building a digital copy of a company’s operations that can detect problems and solve those problems instantly.
Palantir’s financials suggest there is no replacement for the value it provides. It reported accelerating revenue growth over the last year. In the second quarter, revenue grew 48% year over year, compared to 27% in the year-ago quarter.
Moreover, its net income margin was stellar at 33% in Q2, with an adjusted free cash flow margin of 57%. It’s not common for a small software company in the early stages of growth to be reporting margins like Microsoft.
These margins are being driven by high prices that Palantir charges customers. For example, it recently secured a $10 billion contract with the U.S. Army for the next decade. Organizations are willing to pay up for Palantir’s software because the savings realized are that big. Palantir is saving enterprises millions, even hundreds of millions in costs in some cases, providing an attractive return on investment that is driving the company’s growth.
Palantir stock is expensive, trading at high multiples of sales and earnings. But this is a unique software company with a huge opportunity ahead. CEO Alex Karp is aiming to grow revenue by 10x over time, which would bring annual revenue to more than $40 billion from this year’s analyst estimate of $4.1 billion. Based on its current margins, that could equate to $20 billion in annual free cash flow over the long term. Applying a high-growth multiple of 50 to that would put the stock’s market cap at $1 trillion.
2. Advanced Micro Devices
For AI to keep advancing and transform how people work and communicate, it needs more powerful chips. Nvidia has been the biggest winner so far, but investors shouldn’t overlook Advanced Micro Devices (AMD 1.91%). It is the second-leading supplier of graphics processing units (GPUs), and it could be well positioned to meet growing demand in edge computing and AI inferencing that could send the stock from its current $250 billion market cap to $1 trillion.
As AI proliferates across the economy, people will be able to use powerful AI applications and processing on their devices, which makes edge computing a large opportunity for AMD. The company offers a range of high-performance and energy-efficient chips that are aimed at running AI devices and PCs, positioning it to benefit from a booming market estimated to be worth $327 billion by 2033, according to Grand View Research.
Investors were disappointed by the company’s Q2 data center growth of 14% year over year, but management expects stronger demand once it launches its Instinct MI350 series of GPUs. As it continues to bring new solutions to the data center market, AMD’s data center business should accelerate.
AMD’s chips are clearly addressing needs in the AI market. It announced a partnership with Saudi Arabia’s Humain to build AI infrastructure using AMD’s GPUs and software. Meanwhile, Oracle is building a massive AI compute cluster using multiple AMD chips. AMD says it is also working with governments globally to build sovereign AI infrastructure.
Analysts expect AMD‘s earnings to grow at an annualized rate of 30% over the next several years. Against those prospects, the stock trades at a reasonable forward price-to-earnings multiple of 40. There is enough earnings growth here to potentially triple the stock in five years, putting it easily within striking distance of reaching $1 trillion within the next decade.
John Ballard has positions in Advanced Micro Devices, Nvidia, and Palantir Technologies. The Motley Fool has positions in and recommends Advanced Micro Devices, Microsoft, Nvidia, Oracle, and Palantir Technologies. The Motley Fool recommends the following options: long January 2026 $395 calls on Microsoft and short January 2026 $405 calls on Microsoft. The Motley Fool has a disclosure policy.
AI Insights
Kazakhstan establishes Ministry of Artificial Intelligence to spearhead digital nation transformation

Kazakhstan has announced the creation of a Ministry of Artificial Intelligence (AI) and a systemic shift towards a digital state, set to be realised within the next three years. President Kassym-Jomart Tokayev outlined a comprehensive reform plan, highlighting AI as the central driver for transformation across all sectors, from government administration and industry to agriculture and education.
The initiative includes the integration of a digital tenge into the budgetary system. This is reported by the
official website of Kazakhstan’s president.
A key component of this new development phase is the creation of a Digital Code, designed to standardise regulations surrounding technologies, digital platforms, data, and AI.
The Code will serve as the foundational legal framework for both business and government. The establishment of the Ministry of Artificial Intelligence and Digital Development is an institutional step.
AI integration will encompass all spheres, from the economy and industry to public administration and the social sector. Government services are slated to transition to intelligent platforms, while businesses will be encouraged to adopt digital technologies to enhance productivity and competitiveness.
The initiative includes a social component with the launch of the programme, focused on educating students and schoolchildren in the fundamentals of artificial intelligence. Plans are also in place to introduce AI as a separate subject in school curricula for the first time.
Photo: Myvector /
iStock
AI Insights
2 Popular AI Stocks to Sell Before They Fall 46% and 73%, According to Wall Street Analysts

Popular artificial intelligence (AI) stocks Palantir and Arm may be headed for colossal losses.
Shares of Palantir Technologies (PLTR 4.14%) have returned 2,570% since the artificial intelligence (AI) boom began in earnest in January 2023. Arm Holdings (ARM -2.62%) did not go public until September 2023, but shares have since advanced 195%. Those gains have left both stocks trading at rich valuations, so much so that certain Wall Street analysts recommend selling.
- Rishi Jaluria at RBC Capital has set a target price of $45 per share for Palantir. That implies 73% downside from its current share price of $171.
- Javier Correonero at Morningstar has set a target price of $80 per share for Arm. That implies 46% downside from its current share price of $150.
Here’s what investors should know about these popular AI stocks.
Image source: Getty Images.
Palantir Technologies: 73% implied downside
Palantir introduced its Artificial Intelligence Platform (AIP) in April 2023. It serves as a large language model organization tool that complements its core data analytics platforms by letting developers integrate generative AI into applications and workflows. The product has been an unmitigated success, such that sales growth has accelerated in eight consecutive quarters.
Palantir’s advantage lies in its unique ontology-based software architecture. In this context, an ontology is a framework that integrates an organization’s data, assets, and actions into a digital twin that supports decision-making. It also captures the outcome of every decision and feeds the information back into the models, which creates a feedback loop that leads to better insights over time.
International Data Corp. ranked Palantir as the market leader in decision intelligence platforms last year. That bodes well for the company. Grand View Research estimates that data analytics software sales will increase at 29% annually through 2030. “The main factors propelling the data analytics industry expansion are the growing adoption of machine learning and artificial intelligence,” according to the report.
However, Palantir is one of the most richly valued software stocks in history. It currently trades at 126 times sales, which makes it the most expensive stock in the S&P 500 by a long shot. The second-most expensive stock is Texas Pacific Land at 29 times sales. That means Palantir would still be the most expensive stock in the index even if it lost 75% of its value.
In that context, it is entirely plausible that Palantir will suffer a major meltdown at some point in the future. Prospective investors should avoid the stock or, at the very least, keep any positions very small. Current shareholders with a substantial percentage of their portfolios invested in Palantir should consider trimming their positions.
Arm Holdings: 46% implied downside
Arm has long dominated the market for mobile device processors due to its power-efficient architecture. Its central processing units (CPUs) are found in 99% of smartphones. But that quality, coupled with the flexibility of its licensing model — Arm does not make chips, but rather licenses blueprints to customers who develop custom chips — has also helped it gain market share in data centers.
Major technology companies, such as Alphabet, Amazon, Apple, and Microsoft, have designed Arm-based server processors. And Nvidia‘s Grace Blackwell Superchip pairs two Blackwell GPUs with an Arm-based Grace CPU. In total, Arm has added about 10 percentage points of market share in data centers in the last two years, while Intel has lost about 16 points. AMD has also gained share, which accounts for the difference.
That trend is likely to continue as companies look to curb operating costs associated with AI infrastructure by deploying more power-efficient server processors. CEO Rene Hass recently said AI is “driving unprecedented demand for compute that’s not only performant, but also energy efficient. And Arm is the only compute platform built to deliver.”
However, Arm currently trades at 94 times adjusted earnings. That is particularly expensive for a company whose earnings are forecasted to increase at 23% annually through fiscal 2027. Those figures give Arm a price/earnings-to-growth (PEG) ratio above 4, which is traditionally seen as overvalued. Moreover, Arm trades at 39 times sales, which makes it the third-most expensive stock in the Nasdaq-100, behind Palantir and Strategy.
I doubt Arm shares will decline 46% unless the broader market drops sharply, but the stock is very expensive. Investors should wait for a better entry point before putting money into this semiconductor company. Personally, I would feel more comfortable buying at $120 per share, though the valuation would still be stretched even at that price.
Trevor Jennewine has positions in Amazon, Nvidia, and Palantir Technologies. The Motley Fool has positions in and recommends Advanced Micro Devices, Alphabet, Amazon, Apple, Intel, Microsoft, Nvidia, and Palantir Technologies. The Motley Fool recommends the following options: long January 2026 $395 calls on Microsoft, short January 2026 $405 calls on Microsoft, and short November 2025 $21 puts on Intel. The Motley Fool has a disclosure policy.
-
Business2 weeks ago
The Guardian view on Trump and the Fed: independence is no substitute for accountability | Editorial
-
Tools & Platforms1 month ago
Building Trust in Military AI Starts with Opening the Black Box – War on the Rocks
-
Ethics & Policy2 months ago
SDAIA Supports Saudi Arabia’s Leadership in Shaping Global AI Ethics, Policy, and Research – وكالة الأنباء السعودية
-
Events & Conferences4 months ago
Journey to 1000 models: Scaling Instagram’s recommendation system
-
Jobs & Careers3 months ago
Mumbai-based Perplexity Alternative Has 60k+ Users Without Funding
-
Podcasts & Talks2 months ago
Happy 4th of July! 🎆 Made with Veo 3 in Gemini
-
Education2 months ago
VEX Robotics launches AI-powered classroom robotics system
-
Education2 months ago
Macron says UK and France have duty to tackle illegal migration ‘with humanity, solidarity and firmness’ – UK politics live | Politics
-
Podcasts & Talks2 months ago
OpenAI 🤝 @teamganassi
-
Funding & Business3 months ago
Kayak and Expedia race to build AI travel agents that turn social posts into itineraries