Skip to main content
Mallory
← Back to News
privacy-surveillance-policy

DeepMind Uses AI to Formally Verify Parts of Signal’s One-to-One Encryption

Regulatory & Policy · 7 sources · Updated 3h ago · First seen Nov 15, 2026

DeepMind Uses AI to Formally Verify Parts of Signal’s One-to-One Encryption

Google DeepMind researchers used agentic AI to produce computer-checked proofs of selected cryptographic security properties in Signal’s libsignal library. The four-week proof of concept reportedly used 80 AI agents and the AlphaProof Code framework. It covered one-to-one session establishment and message-key updates in compiled AArch64 Linux code from libsignal v0.105.4. The results are not a security certification of the entire Signal application: group conversations, underlying cryptographic implementations and several other components were outside scope, and the proofs depend on stated assumptions about cryptographic primitives.

Separately, Signal disclosed an October US grand jury subpoena seeking information associated with four phone numbers. Only one had a registered account, for which Signal supplied an account-creation timestamp and last-connection date. The disclosure illustrates how limited server-side retention complements message encryption, while the verification work provides additional assurance only for the components examined. Users do not need to change settings in response to the findings. For organizations relying on Signal, independent reproduction and broader verification remain outstanding; the proofs should not be interpreted as eliminating endpoint compromise or other risks outside their scope.

What this changes for youSee the reporting duties and controls this puts on the clock.
Get the context
Event timeline

How this story unfolded

Events from the most recent confirmed update back to the earliest known activity.

14 events · 2016 – Nov 11, 2026
  1. Reported Nov 15, 20264d ago

    OpenAI introduces Control Room for intervening in AI research tasks

    OpenAI introduced Control Room to check research-task requests during inference and block them when problems are detected, targeting intervention within 35 seconds rather than guaranteeing that stopping time. At launch, it covered frontier-model research rather than publicly accessible models.

    Microsoft bloque des mails sans prévenir, Google fait prouver la sécurité de Signal par 80 agents, OpenAI installe un bouton stop sur ses IA : on vous raconte la semaine Cyberguerre - Numerama
  2. Reported Nov 15, 20264d ago

    OpenAI experiences lengthy delays stopping AI research tasks

    OpenAI experienced incidents in which stopping a research task took nearly two hours and 45 minutes. Those delays motivated the development of additional intervention controls.

    Microsoft bloque des mails sans prévenir, Google fait prouver la sécurité de Signal par 80 agents, OpenAI installe un bouton stop sur ses IA : on vous raconte la semaine Cyberguerre - Numerama
  3. Nov 11, 20269d ago

    Signal proof update verifies 10 more routines and reduces trusted hooks

    A merged pull request extended direct verification to 10 additional library routines, including memory operations, UTF-8 handling, time operations, and secret-memory wiping, reducing trusted hooks from 37 to 27. The headline theorem statements remained unchanged.

    What DeepMind's Signal Proof Actually Covers | MeetCyber
  4. Nov 10, 202610d ago

    DeepMind announces AI-generated formal proofs for selected libsignal components

    Google DeepMind announced computer-checked proofs for selected one-to-one session establishment and message-key updates in libsignal v0.105.4, using AlphaProof Code and 80 AI agents, and released the specifications and proofs as open source. The work established specified security properties under stated assumptions, not the security of the entire Signal application, group chats, or underlying cryptographic implementations.

    Microsoft bloque des mails sans prévenir, Google fait prouver la sécurité de Signal par 80 agents, OpenAI installe un bouton stop sur ses IA : on vous raconte la semaine Cyberguerre - Numerama
  5. Nov 5, 202615d ago

    Microsoft mistakenly rejects email from smaller domains without DMARC

    Microsoft began intermittently rejecting messages from smaller sending domains without DMARC records, affecting Outlook, Hotmail, Live, and MSN recipients beyond its announced high-volume threshold. MailRoute reported that the erroneous rejections occurred between November 5 and 9.

    Microsoft bloque des mails sans prévenir, Google fait prouver la sécurité de Signal par 80 agents, OpenAI installe un bouton stop sur ses IA : on vous raconte la semaine Cyberguerre - Numerama
  6. Nov 3, 202617d ago

    Signal publishes the October subpoena and response

    Signal published the subpoena documents after redacting phone numbers, the requested time period, and government personnel identities. The disclosure demonstrated the limited subscriber information available from Signal.

    Machine-checked proofs for Signal's one-to-one chats and a US grand jury subpoena | anoni.net News
  7. Oct 6, 20261mo ago

    ACLU responds to the Signal subpoena with limited account metadata

    The ACLU responded for Signal, stating that three requested numbers had no Signal account. For the sole registered account, Signal supplied its creation timestamp and last connection date.

    Machine-checked proofs for Signal's one-to-one chats and a US grand jury subpoena | anoni.net News
  8. Oct 5, 20262mo ago

    US Attorney's Office subpoenas Signal for information on four phone numbers

    The US Attorney's Office in Washington, DC, issued a subpoena seeking subscriber information associated with four phone numbers. The subpoena carried no nondisclosure order.

    Machine-checked proofs for Signal's one-to-one chats and a US grand jury subpoena | anoni.net News
  9. Reported Jun 25, 20265mo ago

    Microsoft moves to rejecting noncompliant high-volume email

    Microsoft began rejecting noncompliant high-volume email rather than merely routing it to Junk. The reported rejection response was '550 5.7.515 Access denied, sending domain does not meet the required authentication level.'

    SPF, DKIM, and DMARC Are No Longer Optional | MailRoute
  10. Reported Jun 25, 20265mo ago

    Microsoft rolls back the faulty spam rule behind smaller-domain rejections

    Microsoft attributed the erroneous rejection of messages from smaller domains without DMARC records to a faulty spam rule update and rolled it back.

    SPF, DKIM, and DMARC Are No Longer Optional | MailRoute
  11. May 2025

    Microsoft introduces authentication requirements for high-volume email senders

    Microsoft introduced authentication requirements for senders exceeding 5,000 messages per day to its consumer email services, including Outlook.com, Hotmail, and Live.

    SPF, DKIM, and DMARC Are No Longer Optional | MailRoute
  12. 2025

    US court issues an earlier subpoena concerning 37 Signal accounts

    An earlier subpoena concerned 37 accounts. Signal ultimately supplied data for six; seven accounts did not exist, and 24 had no relevant data for the requested period.

    Machine-checked proofs for Signal's one-to-one chats and a US grand jury subpoena | anoni.net News
  13. Feb 2024

    Google and Yahoo introduce bulk-sender authentication requirements

    Google and Yahoo began introducing stronger authentication requirements for bulk email senders, including SPF, DKIM, DMARC, and domain alignment.

    SPF, DKIM, and DMARC Are No Longer Optional | MailRoute
  14. 2016

    Signal publishes its first subpoena response

    Signal's first published subpoena response identified account creation time and last connection date as the two metadata items it retained.

    Machine-checked proofs for Signal's one-to-one chats and a US grand jury subpoena | anoni.net News
Linked entities

Related entities

Vulnerabilities, threat actors, malware, products, organizations, and breaches Mallory has linked to this story.

Open in app →

Affected products

1 linked
Signal

Organizations

1 linked
Signal Messenger
Do you run this?See what this changes for your reporting obligations and which controls it puts on the clock.
Get the Regulatory Context
Mallory platform

See the full picture, correlated to your attack surface.

This page covers what’s public. Mallory adds the parts that aren’t: which of your assets are affected, which threat actors are using it right now, and what to deploy next.

Exposure mappingMap indicators from this story to your assets and identify affected systems in minutes.
Threat actor evidenceEvery observed campaign, victim, and pivot linked to actors named in this story.
Associated malwareMalware, exploits, and IOCs connected to the activity described here.
Detection signaturesYARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Scheduled alertsGet matching new stories delivered to your team as they break, not the next morning.
AI threadsAsk questions about this story and take action on the answers.