Anthropic said infostealer malware copied active Claude browser session cookies from infected computers, letting attackers replay authenticated sessions and consume paid usage without the password or a new MFA code. The attack exploited a valid browser session rather than breaking into an account through its login s...
Published byEdited with GPT-5.6 Luna
Research answer

Create a landscape editorial hero image for this Studio Global article: What did Anthropic disclose about the infostealer-malware campaign that hijacked active Claude login sessions—including how attackers stole. Article summary: Anthropic said a bad actor used common infostealer malware on users’ computers to copy active Claude browser-session cookies, then replay those already authenticated sessions to enter accounts and consume paid Claude usa. Topic tags: general, general web, user generated. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fak
Anthropic warned some Claude users that common infostealer malware had stolen active browser sessions from their computers. Attackers could then replay those sessions to access Claude accounts and consume paid usage without obtaining the account password or passing a new two-factor-authentication check. 18
19
21
The immediate lesson is broader than Claude: once malware takes an authenticated browser session, controls that protect the next login may not stop the attacker already holding that session. Account recovery therefore needs to be paired with endpoint cleanup.
Infostealers are designed to collect data from an infected computer, including browser cookies, saved credentials, and other locally stored authentication information. In this campaign, the stolen data included active Claude login sessions. The attacker replayed the session data as if they were continuing an already authenticated visit, rather than signing in with a newly entered password. 19
22
26
That distinction explains why the attack could avoid a password prompt or a new two-factor-authentication code. MFA can protect a fresh authentication event, but it does not necessarily invalidate a session token that has already been stolen. Reporting on the incident described the technique as bypassing password and MFA checks because the attacker used the valid session itself. 19
23
26
Anthropic’s systems identified suspicious activity associated with affected accounts. One visible warning sign was Claude usage being consumed in a way that did not match the account holder’s activity, including unexpected pressure on usage limits. 18
21
That signal matters because session theft may not look like a conventional account takeover. A user can still have the correct password, while an attacker operates through a session that was copied from the user’s browser. Unexpected usage, unfamiliar activity, or an account notification can therefore be evidence of a compromised device—not necessarily proof that the service’s own login database was breached.
Reports identified five Windows infostealers—Vidar, LummaC2, StealC, RedLine, and Acreed—and Atomic Stealer on macOS. 19
26
These are general-purpose information-stealing tools, not malware described as being built into Claude. The reported explanation was that the malware was already running on users’ infected Windows PCs or Macs and collected Claude’s browser-resident session data along with other credentials and cookies. 20
22
25
The available reporting focused on desktop and laptop computers. It did not establish that phones or tablets were affected by this particular browser-session theft mechanism, but that observation should not be treated as a blanket guarantee that mobile devices are immune to every account or malware threat. 22
Anthropic signed suspected users out of the affected Claude sessions, removed saved payment methods, and said it would refund charges it identified as unauthorized. These steps can cut off the stolen session and reduce the risk of additional paid usage or billing abuse. 17
18
20
21
They do not, however, remove the infostealer from the computer. If the device remains infected, an attacker may be able to steal newly created sessions, passwords, or other browser data after the user logs in again. Endpoint investigation and cleanup therefore remain essential. 20
If Anthropic notified you, or if your account shows unexplained usage, treat the computer used to access Claude as potentially compromised.
Forced logouts and card removal are useful containment measures, but they are not a substitute for cleaning the endpoint. The core risk in this incident was the infected computer’s ability to expose authenticated browser data in the first place. 20
25
Studio Global AI
This page includes a source-backed answer you can continue inside Studio Global.
Anthropic said infostealer malware copied active Claude browser session cookies from infected computers, letting attackers replay authenticated sessions and consume paid usage without the password or a new MFA code.
Anthropic said infostealer malware copied active Claude browser session cookies from infected computers, letting attackers replay authenticated sessions and consume paid usage without the password or a new MFA code. The attack exploited a valid browser session rather than breaking into an account through its login screen, so changing MFA settings alone may not remove the underlying risk.
Anthropic signed out suspected sessions, removed saved payment methods, and identified refunds for unauthorized charges—but those account actions do not clean an infected computer.
Anthropic said infostealer malware copied active Claude browser session cookies from infected computers, letting attackers replay authenticated sessions and consume paid usage without the password or a new MFA code. The attack exploited a valid browser session rather than breaking into an account through its login s...
Published byEdited with GPT-5.6 Luna
Research answer

Create a landscape editorial hero image for this Studio Global article: What did Anthropic disclose about the infostealer-malware campaign that hijacked active Claude login sessions—including how attackers stole. Article summary: Anthropic said a bad actor used common infostealer malware on users’ computers to copy active Claude browser-session cookies, then replay those already authenticated sessions to enter accounts and consume paid Claude usa. Topic tags: general, general web, user generated. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fak
Anthropic warned some Claude users that common infostealer malware had stolen active browser sessions from their computers. Attackers could then replay those sessions to access Claude accounts and consume paid usage without obtaining the account password or passing a new two-factor-authentication check. 18
19
21
The immediate lesson is broader than Claude: once malware takes an authenticated browser session, controls that protect the next login may not stop the attacker already holding that session. Account recovery therefore needs to be paired with endpoint cleanup.
Infostealers are designed to collect data from an infected computer, including browser cookies, saved credentials, and other locally stored authentication information. In this campaign, the stolen data included active Claude login sessions. The attacker replayed the session data as if they were continuing an already authenticated visit, rather than signing in with a newly entered password. 19
22
26
That distinction explains why the attack could avoid a password prompt or a new two-factor-authentication code. MFA can protect a fresh authentication event, but it does not necessarily invalidate a session token that has already been stolen. Reporting on the incident described the technique as bypassing password and MFA checks because the attacker used the valid session itself. 19
23
26
Anthropic’s systems identified suspicious activity associated with affected accounts. One visible warning sign was Claude usage being consumed in a way that did not match the account holder’s activity, including unexpected pressure on usage limits. 18
21
That signal matters because session theft may not look like a conventional account takeover. A user can still have the correct password, while an attacker operates through a session that was copied from the user’s browser. Unexpected usage, unfamiliar activity, or an account notification can therefore be evidence of a compromised device—not necessarily proof that the service’s own login database was breached.
Reports identified five Windows infostealers—Vidar, LummaC2, StealC, RedLine, and Acreed—and Atomic Stealer on macOS. 19
26
These are general-purpose information-stealing tools, not malware described as being built into Claude. The reported explanation was that the malware was already running on users’ infected Windows PCs or Macs and collected Claude’s browser-resident session data along with other credentials and cookies. 20
22
25
The available reporting focused on desktop and laptop computers. It did not establish that phones or tablets were affected by this particular browser-session theft mechanism, but that observation should not be treated as a blanket guarantee that mobile devices are immune to every account or malware threat. 22
Anthropic signed suspected users out of the affected Claude sessions, removed saved payment methods, and said it would refund charges it identified as unauthorized. These steps can cut off the stolen session and reduce the risk of additional paid usage or billing abuse. 17
18
20
21
They do not, however, remove the infostealer from the computer. If the device remains infected, an attacker may be able to steal newly created sessions, passwords, or other browser data after the user logs in again. Endpoint investigation and cleanup therefore remain essential. 20
If Anthropic notified you, or if your account shows unexplained usage, treat the computer used to access Claude as potentially compromised.
Forced logouts and card removal are useful containment measures, but they are not a substitute for cleaning the endpoint. The core risk in this incident was the infected computer’s ability to expose authenticated browser data in the first place. 20
25
Studio Global AI
This page includes a source-backed answer you can continue inside Studio Global.
Anthropic said infostealer malware copied active Claude browser session cookies from infected computers, letting attackers replay authenticated sessions and consume paid usage without the password or a new MFA code.
Anthropic said infostealer malware copied active Claude browser session cookies from infected computers, letting attackers replay authenticated sessions and consume paid usage without the password or a new MFA code. The attack exploited a valid browser session rather than breaking into an account through its login screen, so changing MFA settings alone may not remove the underlying risk.
Anthropic signed out suspected sessions, removed saved payment methods, and identified refunds for unauthorized charges—but those account actions do not clean an infected computer.