Resemble AI

Real-time, multimodal deepfake detection and watermarking for enterprise. Verify identity, meet AI compliance standards — via API or on-prem.

[Resemble AI @resembleai]

🌍 DETECT-World is live.

Our third generation of deepfake detection and the first built on world model architecture.

Benchmarked across 250+ generative models:
• 99.5% audio
• 95.8% image
• 98.2% video

Now powering Resemble Detect via streaming + batch API. 👉 https://t.co/AnIcG2xXSl

[Zohaib Ahmed @zohaibahmed]

Today we’re officially launching Resemble Agent Detection, a System One model that reads behavior to tell you whether a human, bot or AI agent is on your site or in your app.

Schwab and JPMorgan are already blocking AI agents from customer accounts over liability. You know they thought through the opportunity cost, because agents are starting to transact on behalf of real customers.

[Tasnia Huque @tasniahuque]

“Camera on” is becoming a weaker proof of identity.

Companies like Tavus are making lifelike AI participants possible: faces, voices and conversations that respond in real time. Incredible possibilities, but also growing the need to know who you are "actually" talking to.

Resemble AI documented 821 verified deepfake attacks and 15,736 victims in H1 2026 alone. This is why @resembleai + Microsoft Teams integration is so epic: flag suspected impersonators right where they’re likely to show up. In your interviews, executive calls and vendor meetings. Detect suspicious AI-generated audio and video live, without requiring an extra bot in the call. Seamlessly embedded in a product experience that you already heavily use.

https://t.co/giPzzqGZWA

[Matt Ridenour @matthewridenour]

@GoogleStartups x @GoogleDeepMind SFTECHWEEK event is next level vibes!!

Thanks everyone for coming out for a full house talking research to reality, awesome panel with @o_lacombe, @liu8in - @HeyGen and @zohaibahmed - @resembleai

Plus 5 amazing Gemini startup demos from @krea_ai, @raindrop_ai @benhylak, @mathgpt_app @ZeroCater

[Resemble AI @resembleai]

@realManojGarg @SayantikaSays @nitin_gadkari @narendramodi @BJP4India @resemble_detect is this ai?

[Resemble AI @resembleai]

@tearex001 @realManojGarg @SayantikaSays @nitin_gadkari @narendramodi @BJP4India calling in our detection bot to answer. @resemble_detect is this ai?

[Resemble AI @resembleai]

Tomorrow our Head of JevRel, @0xDevShah , will be speaking at @FeatherlessAI and @Tiny_Fish's Hack and Demo night. He'll be showing how we do agent detection, kind of a must have when everyone is using agents.

Super hot topic x smart speakers = amazing event. See you there > https://t.co/ENDSBSBKI0

[Dr. Alex Wissner-Gross @alexwg]

My conversation with Bradley Olson of @Axios, @ZohaibAhmed of @ResembleAI, and @NoelleRussell_ on AI agent safety at @ImaginationXYZ @Google is out. https://t.co/b2Pp0CoxtL

[Resemble AI @resembleai]

@zohaibahmed

[Zohaib Ahmed @zohaibahmed]

The line between a bad joke and a geopolitical crisis is razor thin. And causing one could cost under $200.

Investigative journalists @christogrozev and @tashurkevich proved it when they hired a freelancer on a Russian gig marketplace to create a video of them with a NATO general, chatting on a yacht. 

The general is fake - not just synthetic - but actually nonexistent. The freelancer who agreed to produce the video never asked why they wanted it. The video took an hour and cost roughly $80. Someone placed it for another $80, and six Russian news sites ran it as news within an hour.

Influence operations used to require intel or power or serious capital. Now anyone with a credit card and free time can put hyperrealistic fakes in front of millions.

The US midterms are five weeks out. This is a dangerous period where synthetic media can do a ton of damage, because a clip only needs to catch the algorithm right once and it can rack up millions of views in days. Corrections almost never reach the same audience.

Verifying political video and audio before sharing it takes seconds. It should be a habit for anyone voting in an upcoming election, and our free Chrome extension lets you check media directly in your feed.

Watch the investigation here > https://t.co/Cvep2mLxhG

[Resemble AI @resembleai]

@0xDevShah Agent Detection for voice is live and easy to integrate with LiveKit, Pipecat, and your personal WebRTC stack.

Try it at https://t.co/XIfOWkByBk

[Dev Shah @0xDevShah]

Today, we're extending the Agent Detection-1 family of foundational models to secure your Voice agents. Your voice agent can now tell whether it's talking to a real person or another AI.

Voice agents are already on the phone. They book appointments, confirm reservations, and handle support calls, and more and more often, the one picking up is also a voice agent.

When two voice agents meet, neither knows it's talking to an AI. So they keep talking for hours, sometimes days, and every minute burns speech-to-text, text-to-speech, and LLM spend.

We want your voice agent to know who's on the line, so it can adapt the moment the call connects.

Agent Detection-1 Voice tells you who's calling the moment they speak:
> Real person or AI, with 99% accuracy
> So your agent can simply ask, skip the small talk, finish the job, and hang up
> real people still get the full conversation
> every call gets logged on your Resemble dashboard

You can integrate our models across your Voice agent stack, including @livekit and @pipecat_ai!!

One model family, for your web platform and your phone lines. My DMs are open for more details!

[Dev Shah @0xDevShah]

more than 50% of your web traffic is just AI agents. if you block them all, you’re literally losing real human customers behind these agents. so don't block them all, identify them and then decide which ones to block and which ones to send to an MCP!!

dms open. I'll help you set up Agent Detection in mixpanel, posthog, amplitude, datadog, or whatever analytics stack you're using.

[Dev Shah @0xDevShah]

Remember when every website rushed to get mobile friendly so they wouldn't lose half their users? Well, now it's about making your website agent friendly!

Personal agents are out there browsing, buying, and getting stuff done for real people. If you just block them all, you're losing a chunk of your revenue. Instead, identify the helpful agents and give them an agent-friendly surface to do their thing.

[Dev Shah @0xDevShah]

More than half of customer traffic is literally just agent swarms now. goal is not to block them. If you can identify them and route the personal assistant agents to agent-friendly interfaces like MCP and llms.txt, or where they can transact safely, and only prune the malicious traffic.

[Indra @IndraVahan]

dude this is so cool. imo the future of the internet is every website split into two: one serves the human and another serves the agent

[Dev Shah @0xDevShah]

Today, we are launching Agent Detection-1, the first system-one model built to classify visitors on your web platform.

Agents are already on your web platform. Muse, Instinct, Claude Code, Codex, and thousands more are booking, buying, and filling out forms. Behind each agent could be a real customer.

Cloudflare, Akamai, and other bot tools are built to block agents. Analytics like Amplitude, Datadog, and Google Analytics can't tell them apart from people.

We want to detect agents, figure out which ones are useful, and redirect them to where they can actually get things done.

Agent Detection-1 tells you who's at the door, the moment they arrive:
> Person or agent, and which agent
> Send useful agents to your MCP server, llms.txt, or agent card
> Stop the bad ones, like scrapers and credential stuffers
> Every visit gets logged on your Resemble dashboard

Just add a one-line script and find out who's visiting your website. Starting at just $1 per 1,000 visitors.

[Zohaib Ahmed @zohaibahmed]

Introducing our AI Agent Detection model, the first System One model built for Agentic Detection! 🎉

Meta's Muse is beating ChatGPT's early adoption: 1.8 million downloads in its first 12 days across the US and Canada, vs 1.3 million for ChatGPT.

Everyone has figured out what an LLM is and how to use it. The next productivity jump is letting one act on your behalf, starting with booking travel, buying groceries, and filling out the forms you don't want to spend time on. Amazon has already blocked Muse from shopping on its platform, can’t blame em.

[Resemble AI @resembleai]

@costa__io Might be time to add deepfake detection to the interview process

[Telnyx @telnyx]

How do you detect an AI-generated voice on a live call?
We wrote the answer with the @resembleai team.

Post-call review is too late.
By the time anyone listens to the recording, the credentials are already gone.

Detection has to happen while the call is still live.
The article covers how detection works in production, from in-call streaming to post-call audit, and what a real phone call does to accuracy.

Resemble tested this across real PSTN codecs, and published the numbers as they came out.

3 of 5 codecs hit 100% accuracy on both real and synthetic audio.
G.728 and iLBC, the two most compressed, didn't.

Read more about it here: https://t.co/bPMYfvuwrP