Senior Product Director @zama. Co-founder @cheqd_io. Co-chair of Technical SteerCo @DecentralizedID. Ex @FinTechLabLDN, @inside_r3, @Accenture
Paris, France
Joined October 2008
- Tweets11.6K
- Following1.4K
- Followers4.6K
- Likes41.1K
Ankur Banerjee retweeted
I’m very excited and honored, to announce that I’ve been named Chief Business Officer at @zama .
I’ve spent almost twelve years in crypto. I’ve watched it go from a fringe experiment to something institutions can no longer ignore, through every cycle, every boom, every collapse, every slow rebuild of trust after. One gap never closed through any of it - blockchains ask every institution, every business, every person to run their finances in the open. That was never going to be what brings real money onchain at scale.
I joined Zama in November 2025, right before mainnet, because I believe confidentiality is the infrastructure that finally makes that possible - Zama makes it possible. It’s the bridge that lets the institutions holding the world’s capital actually move onchain.
This past year our business team signed partnerships across major players, launched confidential DeFi, and built a pipeline that holds up against anything else in this space. That work has been a game-changer putting Zama at the forefront, showcasing what confidentiality can bring to the table and what the new internet of tomorrow could look like at scale.
@randhindi and the board are putting real weight behind where this team can take Zama next. I feel that responsibility.
Thank you to Rand, to the board and to every person at Zama who makes this real every day.
Now, let's go win!
Such a weird parallel universe where a lot of the attendees of the recent White House dinner (Nadella, Zuckerberg etc) have actually started using the term “super intelligence”
Facts 👇🏽
If ECDSA and other traditional cryptography is broken, the first to fall (with impact) would be badly secured government services or banks that are laggards on technology (not necessarily all legacy banks, just the ones bad at tech).
As for blockchains, I hope the industry push the frontier to using even more secure forms of cryptography.
A few people yesterday asked how AI could impact FHE / @zama, so here is a quick tldr:
1. There is no evidence that lattices / FHE can be broken by AI, SI or quantum. If anything, there is a growing consensus that fully breaking lattices might imply P = NP.
2. However, it's almost certain that better attacks will be discovered, which would require increasing parameter size and taking a short-term hit on performance. The Zama protocol is modular, so we can increase parameters size or swap out any component quickly, without changing the UX or rotating user keys.
3. Zama uses FHE only for privacy, so breaking FHE would not allow stealing funds that are shielded, it would only allow seeing balances and amounts. This is unlike ZK, where breaking the cryptography means you can steal funds and mint infinite tokens.
4. We do a lot of work on cryptanalysis and improving the security of FHE. We spend millions a year on audits (AI and human), research grants, bounties, have 30 PhDs in the team, etc. We have a plan A (the current thing we do), but also a plan B and a plan C if things go wrong. Whatever the future throws at us, we will have a way to adapt to offer the same service.
5. Hashes have been broken in the past, many times (MD5, SHA1, etc), so saying they are more secure than lattices, which have been extensively studied for decades is a stretch to say the least. Not saying its impossible, but history is on the side of lattices here.
Lets also remember that if indeed lattices and elliptic curves are broken and there is no public-key encryption anymore, then hiding your balances onchain is the least of your worries, because the world as we know it would be over: no private communications, no online banking, your AI prompts would be public, etc.
Someone at @TGVINOUI is clearly confused about what the show Modern Family is about 😅
Very cool implementation of what's possible with Privacy Pass!
This, and many other approaches, is a whole ecosystem of solutions for agentic ID that we need.
The internet needs proof of humanness more than ever. We have a unique opportunity to build it with privacy at the core.
As bots become harder to stop and identity theft grows more sophisticated, platforms increasingly need stronger ways to verify who’s real. But stronger verification shouldn’t have to mean collecting more personal data.
We should be asking not just what data is needed, but also what data isn’t.
That’s what the team at @withpersona is working toward with Relay.
Relay uses Privacy Pass, an open IETF standard, to let organizations verify claims like humanness or age without learning your identity, all while preserving the best-in-class accuracy of Persona’s platform.
Proud of the team for the work on this one. Excited to see where it goes
Ankur Banerjee retweeted
Correction: Trillions of dollars in assets will be Tokenized once Confidential tokenization is here
Ankur Banerjee retweeted
In November 2025, I left a job I loved to join a promising privacy project that needed someone to lead their ecosystem and growth strategy. They were not live yet, but had super ambitious plans and tech most didn't believe actually worked.
Sep 2026, Zama is now the fastest growing Privacy protocol, with the most credible names in the space building on it and a truly inspiring roadmap.
We are still early, still bumps ahead.
But - I have never been more convinced that November 2025 I made one of the most impactful decisions of my life, right after marrying my husband and having my son.
Zama will be everywhere.
Confidentiality goes wider for onchain finance.
Why would you NOT want better privacy that what banks/fintechs can offer?
Ankur Banerjee retweeted
Visa's stablecoin settlement just hit a $20B run rate. Up 15x in a year.
The rest of Visa grew 10% over the same period.
Yes, this is a small slice of a $4T-a-quarter network
But it's growing 150x faster than the core.
Stablecoin-linked cards are in hypergrowth mode as my podcast cohost @cuysheffield likes to say.
Visa now runs 160+ stablecoin-linked card programs, up nearly 200% in a year.
🤖 Made with AI
Have never encountered a website whether picking either option makes an iota of difference to whether I get asked to login next time again or not 🫠
Sadly, they won’t be “sued to oblivion” and the only people making a payday from this would be Experian/Equifax.
At best it results in a class-action that result in less than $5 per person in the breach.
Data needs similar regulations like those enacted after the banking sector after the Great Recession.
I really recommending reading this.
In summary, a company which does ID verification for in-person interactions (hotels, car rentals, ID verification for alcohol or marijuana, etc) has some how exposed over 153,000,000 drivers licenses for people in the United States and Canada.
It is a catastrophic data breach, probably one of the worse I've ever seen. If you're in the United States and have traveled, gotten a hotel, purchased marijuana or alcohol, there is a high probability you're in this.
Unlike other breaches, this includes a photo of the person (from the license), making verification you've identified the person significantly easier.
This poses a significant threat to celebrities (musicians, YouTubers, streamers, adult entertainers, actors, etc), politicians, lawyers, wealthy people (CEOs, investors, people of public interest), Law Enforcement Officers, etc
Krebs himself, and several other security researchers, have already confirmed they're in the data leak.
tl;dr gah damn dawg this company is going to be sued into oblivion
krebsonsecurity.com/2026/09/…
Duolingo’s “Chess Review” feature is finally general availability!
I’ve had this for a couple of months, and it’s genuinely well-executed to show what the better alternatives during a game would have been, to help you start identifying patterns better. 😗🤌🏽
blog.duolingo.com/chess-game…
Truly an experience of a lifetime! Earlier this week, I got to fly on a zero gravity flight, which is run by a subsidiary of CNES (the French national space agency). They take an Airbus A310 plane with test pilots that fly the plane manually in parabolic paths that simulates Martian (1 parabola), Lunar (2 parabolas), and weightlessness (13 parabolas) for 20 seconds. Before and after each period of weightlessness, we had hypergravity where you experience 1.8g (which doesn't feel that bad surprisingly, unless you stand up).
PS - I did not throw up! (Less than 10% of the people who took the flight did.)
@zama_intern The FHE bear went to zero gravity!🧸
Ankur Banerjee retweeted
We are open sourcing Umbra, an isolated cloud environment for AI agents.
github.com/concrete-security…
Umbra came from a internal need at Zama where we wanted to use agents across the company without giving them more access than they actually needed.
Agents can write code, install software, read internal data, use credentials and interact with production systems. As they get more capable, it becomes unrealistic to review everything they do step by step.
So the idea behind Umbra is simple: put the agent in a box. Give it a machine, let it work inside, and focus on what happens at the boundary.
The VM is theirs. The edge is ours.
Anything that touches github, slack, notion, internet, production or any other external system goes through a controlled proxy. That is where we apply policy.
Can this agent access this repo? Is github read-only? Can it push code? Can it reach production? Which credential does it need?
We grant access only when needed, and secrets are injected only for the operation being authorized.
The same applies to what comes out. Code, data or actions can be checked before they leave the environment.
Umbra is still experimental, but the key idea is: give agents enough freedom to be useful, assume they will sometimes make mistakes, and keep tight control over the boundary.
Agents can be unpredictable and most likely will always be. Boundary shouldn't be.
Ankur Banerjee retweeted
I’ve just spent hours responding to an avalanche of comments about the ChatGPT plugin for Apple Messages. Here are the main ones:
💬 "Anyone can screenshot your messages"
Yes. They can also print them and fax them to the FBI. People using end to end encryption accept the risk that another participant can deliberately betray their trust. They don’t expect that person to give third party software ongoing access to read, search and analyse their private conversations.
💬 "I opted in, so it’s not a problem"
You opted in. Everyone you communicate with didn’t.
💬 "The encryption still works, so you’re wrong"
Yes, the cryptography still works. My concern is third party access to the readable content before or after decryption.
💬 "The recipient could always share your messages"
Yes. The recipient could betray your trust. That doesn’t justify automating third party access to your private conversations.
💬 "Any Mac app with Full Disk Access could already do this"
Yes. Apple specifically warns that Full Disk Access can expose Messages. OpenAI has now productised that capability for ordinary people who aren't aware of the potential implications, at scale. I don't think Apple saw this coming.
💬 "This is no different from giving AI access to email, SMS, Slack or Teams"
It is. Those services don’t provide the same end to end encryption model as iMessage. Think Signal, not SMS. This Proton, not Gmail.
The screenshot one is my favourite.
I’ll either update my original article with an FAQ or write a dedicated follow up. Thankfully, these comments represent less than 2% of the responses I’ve received here. I haven’t even looked at LinkedIn yet.
My concern with this implementation is scale. One person can enable it on their Mac and expose accessible conversations involving potentially hundreds of people who never enabled ChatGPT, never agreed to the access and aren’t notified. Depending on their Apple and iCloud configuration, that can include historical conversations, participant identifiers, contact information and attachments.
Proton summarised my argument well after reading the article. I recommend reading their post first, then clicking through to mine for the full analysis.
My Substack is free. I’m building a community around corporate surveillance, government overreach, privacy and personal digital safety.
I’m interested in what the technology can do, what information it can access, who else can be affected and what happens when that capability is deployed at scale.
Recently, OpenAI shipped a plugin that lets ChatGPT read your Mac's Messages app, search it, and draft and send replies.
However, this feature essentially renders Apple's end-to-end encryption obsolete, because your messages are now sitting on OpenAI's servers in plaintext. The plugin needs Full Disk Access, Contacts and Automation. If you have Messages in iCloud switched on, that's potentially years of history, attachments and identities, plus every SMS 2FA code your iPhone forwards to your Mac.
OpenAI says it runs locally and doesn't build an index. Local execution isn't the same as local processing. The moment ChatGPT summarises a thread or drafts a reply using a hosted model, that content leaves the machine.
In 2025 a US court ordered OpenAI to preserve ChatGPT conversations for the NYT case, including ones people had deleted. Under the CLOUD Act, data in OpenAI's control is reachable by US legal process wherever it's stored, and they can be barred from telling you it happened.
Governments have spent a decade demanding a backdoor into iMessage and getting refused. This one installs itself, for free, with a toggle.
If you've enabled it, everyone who has ever texted you is in it, and none of them know.
Source @Paul__Walsh: paulfwalsh.substack.com/p/gi…
Ankur Banerjee retweeted
new blog post on how on-call is essentially just theatre now
boristane.com/blog/on-call-i…
Instinct/Pally feel so refreshing to what OpenClaw should have been. More than half the time with OpenClaw, I find myself debugging it to see why it’s crashed or long loops where the dashboard is just hung and unresponsive.
French banks are so incorrigible. Sometimes for bank transfers, they require PDFs actually printed out and sigend by hand (refuse to accept PDFs that are digitally filled in).
Except...very recently, my bank switched to offering a DocuSign instead. I rejoiced! Progress!
But then...they sat ignoring the request, and when I finally called them, they said they rejected it because it was "digitally signed" and needed to be signed by hand.
No amount of trying to explain to them that the bank itself had switched to DocuSign would convince them otherwise. 🥲