Threat reportPhishingTL-2026-4778
Anthropic-Themed Phishing Campaign (Fake Private Preview Invites via GoTo Webinar) Steals Google and Microsoft Credentials and Bypasses Email Security Gateways
Anthropic-Themed Phishing Campaign (Fake Private Preview (TL-2026-4778) is a medium-severity phishing campaign, first published 2026-10-29. It has no confirmed attribution, affects Google Google account / Workspace sign-in (credential target), maps to 8 MITRE ATT&CK techniques (T1036.005, T1056.003, T1204.001), and is covered by 9 detection rules and 9 indicators of compromise.
- Severity
- MEDIUMAssessed severity
- CVEs
- 0None referenced
- Techniques
- 8MITRE ATT&CK
- Actors
- 0Not attributed
- Detection rules
- 9SPL · KQL · Sigma
- IOCs
- 9Indicators of compromise
Key facts for TL-2026-4778
- Threat ID
- TL-2026-4778
- Severity
- MEDIUM
- Status
- ACTIVE
- Category
- PHISHING
- First published
- Last reviewed
- Attribution confidence
- LOW
- Motivation
- FINANCIAL
- Target sectors
- enterprise, technology, general-business
- Target regions
- Global
- Detection rules
- 9
- Indicators of compromise
- 9
How Anthropic-Themed Phishing Campaign (Fake Private Preview works
Since September 2026 an unidentified actor has emailed enterprise users worldwide with fake invite-only Anthropic AI preview invitations to harvest Google and Microsoft account credentials. Cofense reports the emails passed Proofpoint SEG and Microsoft Defender for Office 365, likely helped by a legitimate GoTo Webinar sender address, a Zapier-shortened link and two CAPTCHA gates.
Cofense researchers (reported 2026-10-29 by CSO Online/Computerworld and in Dark Reading coverage) describe a credential-phishing campaign that has run since September 2026. The lure exploits fear of missing out among enterprise employees by offering exclusive, invite-only access to a preview of Anthropic's AI technology. The sender display name is "Austin from nthropic" (missing the leading 'A'), the email subject contains a grammatical error, and the message is sent from the legitimate address customercare@gotowebinar.com with GoTo Webinar branding, not from an Anthropic domain.
The delivery and evasion chain is low-sophistication but layered. The malicious link in the email is shortened (reported as Zapier URL shortening, with other URL shorteners also referenced). Clicking it first presents a Cloudflare CAPTCHA, which prevents automated URL scanners from reaching the phishing content. The victim then lands on a page spoofing a Zoom meeting/registration invitation that asks for a work email address. Choosing "Continue with Google" or "Continue with Microsoft" triggers a second CAPTCHA (reported to block VPN and Tor users and analysis tools) before a spoofed SSO login page is shown. Reported lookalike hosts include goo.gle-access-eventregister.icu (fake Google login) and office.com-access-sharepoint.com (fake Microsoft login), which use subdomains and dashes to resemble legitimate brands.
Credential capture differs by provider: the spoofed Microsoft page captures both username and password, while the spoofed Google page appears to capture only the username (suspected misconfiguration). No confirmation message or redirect follows submission. Cofense reported the emails reached inboxes through Proofpoint Secure Email Gateway and Microsoft Defender for Office 365, and suggested the legitimate GoTo Webinar sender, the Zapier-shortened link and the two CAPTCHAs as possible explanations; one outlet noted the CSO article did not detail a specific bypass mechanism. The actor is unidentified, no victim counts or geographic breakdown were published, and no CVE is involved. Recommended defenses: verify AI preview invitations through known account contacts, treat inconsistent branding across email, landing page and login as a red flag, deploy phishing-resistant MFA (passkeys), and reset credentials and audit account activity for anyone who submitted them.
Note: a separate, distinct Anthropic-branded campaign reported by Malwarebytes (fake Claude Max giveaway with a browser-in-the-browser Google login, reported 2026-09-23) is not attributed to this campaign and its indicators are not merged here. Source caveat: Dark Reading returned HTTP 403 to direct fetch; details come from search summaries of that article and from CSO Online/Computerworld and aggregator pages summarizing the Cofense research.
MITRE ATT&CK techniques used in TL-2026-4778
Defense Evasion
T1036.005 Match Legitimate Resource Name or Location; T1480 Execution Guardrails; T1684.001 Impersonation
Credential Access
Execution
Initial Access
Resource Development
Reconnaissance
Affected products and versions in Anthropic-Themed Phishing Campaign (Fake Private Preview
Remediation for Anthropic-Themed Phishing Campaign (Fake Private Preview
Immediate actions
- Block goo.gle-access-eventregister.icu and office.com-access-sharepoint.com at DNS/web proxy and search mail logs for delivery of the lure
- Hunt for inbound mail from customercare@gotowebinar.com with display name 'Austin from nthropic' or Anthropic preview-invite subjects, and purge matches
- Reset passwords and revoke sessions/tokens for any user who visited the lookalike pages or submitted credentials, then audit sign-in and mailbox-rule activity
Workarounds
- Add mail-flow rules to quarantine or banner webinar-platform mail with unexpected brand names in the display name
- Detonate or rewrite shortened links with a scanner able to solve or bypass CAPTCHA gates
Longer-term hardening
- Deploy phishing-resistant MFA (passkeys/FIDO2) for Google Workspace and Microsoft 365
- Add detections for mismatched brand (Anthropic) versus sender infrastructure (GoTo Webinar) and for URL-shortener links in webinar/invite mail
- Train users on AI-preview and waitlist social engineering and encourage verifying invitations through known account contacts
Timeline of Anthropic-Themed Phishing Campaign (Fake Private Preview
- Per reporting, the Anthropic-themed phishing campaign began 'since September' 2026 (exact day not stated; first of month used as placeholder), with emails sent worldwide offering fake early access to Anthropic AI technology.
- Malwarebytes reported a separate Anthropic-branded Claude Max giveaway phishing page using a browser-in-the-browser Google login; not attributed to this campaign, listed for context.
- Coverage of the separate Malwarebytes Claude Max phishing campaign published, showing broader Anthropic-brand abuse during the same period.
- Campaign assessed as still active at time of reporting; actor unidentified and no victim counts or regional breakdown published.
- Cofense analysis detailed the flow: shortened link, Cloudflare CAPTCHA, Zoom-spoofed registration page, second CAPTCHA, then spoofed Google (username only) and Microsoft (username and password) login pages.
- Cofense research published via CSO Online/Computerworld: fake invite-only Anthropic preview emails from customercare@gotowebinar.com ('Austin from nthropic') that passed Proofpoint SEG and Microsoft Defender for Office 365.
Sources cited for Anthropic-Themed Phishing Campaign (Fake Private Preview
- Anthropic-Themed Phishing Emails Bypass Top Security Tools (Dark Reading)
- Fake Anthropic AI previews lure in overeager employees (Computerworld)
- Fake Anthropic AI previews lure in overeager employees (CSO Online)
- Phishing campaign impersonates Anthropic preview invites to steal enterprise credentials (Get Ready for Agents)
- Hackers Steal Microsoft Credentials With Fake Anthropic Previews as Google Logins Also Targeted (Zetik)
- Anthropic-Themed Phishing Emails Bypass Top Security Tools (daily.dev)
- Security Alert: Spam Emails Containing Suspicious Links From GoToWebinar Trial Account Users (GoTo, related abuse of the platform)
- Malwarebytes Identifies Claude Max Credential Phishing Campaign (distinct Anthropic-branded campaign, context only)
Detection coverage for TL-2026-4778
As of 2026-10-29, Threadlinqs Intelligence publishes 9 detection rule(s) for TL-2026-4778 across Splunk SPL, Microsoft KQL and Sigma, covering 9 indicator(s) of compromise. The whole corpus is readable without an account; a free account unlocks full detection query text in Splunk SPL, Microsoft KQL and Sigma; paid tiers add raw indicator values, correlation and the MCP server. Threadlinqs MCP server · View plans.