// AI threat desk · 15 Nov 2026
Home/LEAK · Data Leaks
Data LeaksMediumAITOKENSCVSS not assigned

At least a dozen AI developer platforms hit in attacks seeking AI access tokens

At least a dozen AI developer platforms were hit over the past month, including Chainstack and WorkOS, in attacks seeking AI access tokens, SecurityWeek reported.

· 2 min read
AI-generated image of a security analyst monitoring blurred access-token alerts in a dark operations centre.
AI-generated image, not a photo of the event.
Key takeaways
  • SecurityWeek said at least a dozen AI developer platforms were hit over the past month.
  • Chainstack’s accessed database may have exposed customer names, member email addresses and password hashes.
  • WorkOS said hackers accessed sales-application data including business contacts and customer account histories.

The incidents reached different types of customer and business information. Chainstack said an accessed database contained organization names, account-holder names, member email addresses and password hashes. WorkOS said attackers accessed business contact details and summaries of customer communications.

SecurityWeek said the campaign included Andromeda, Brev, CommandCode, Fireworks AI, Lightning AI, Northflank, OpenCode, Railway, Trigger.dev, Warp and Zeabur. The Cloud Security Alliance said attackers were reaching AI-provider credentials through hosting platforms, evaluation services and developer consoles. Wiz published its report on Oct 30, 2026.

On this page

Chainstack and WorkOS disclosed attacks that relied on Artex

Chainstack and WorkOS disclosed attacks that relied on the open-source Artex AI penetration-testing framework for automation. SecurityWeek said the attacks did not appear related, and that initial access varied across incidents, including compromised credentials, third-party dependencies and vulnerability exploitation.

The company said attackers reached a development environment and parts of its production environment through a third-party component. Chainstack said the accessed database contained customer account information, and that node credentials might also have been exposed because attackers reached systems storing them.

Hackers appeared to hunt for AI access tokens

SecurityWeek said the hackers appeared to be hunting for AI access tokens they could monetize and relied on AI agents to speed up and streamline attacks. Wiz said the rise was driven by LLM-jacking and LLM-broker infrastructure, which made AI access tokens monetizable.

Fireworks and Andromeda reported separate credential incidents

Andromeda said it identified unauthorized credential access on October 13, disabled compromised accounts that day and observed no further unauthorized activity. It said the activity began as part of an industry-wide attack around October 9 and affected a limited number of customer accounts.

Wiz said two compromised enterprise credentials had identical first unauthorized activity within 30 minutes of reported credential access within the vendor breach.

Chainstack prompted customers to rotate node credentials. Fireworks told affected customers to replace revoked API keys and update applications that used them.

Wiz observed multiple threat actors using the same compromised keys, which “suggests a robust market for this type of credential”. SecurityWeek said multiple companies linked attacks attributed to the Chinese APT Storm-0940 to use of the Artex framework.

What to do now

  1. Rotate your Chainstack node credentials.
  2. Replace revoked Fireworks API keys, create new keys in your Fireworks account and update the applications that used the revoked keys.

FAQ

What happened in the AI developer platform attacks?

SecurityWeek said hackers targeted at least a dozen AI developer platforms over the past month, seeking AI access tokens they could monetize. It said the attacks did not appear related and used different initial-access methods.

Which AI developer platforms were targeted?

SecurityWeek named Andromeda, Brev, CommandCode, Fireworks AI, Lightning AI, Northflank, OpenCode, Railway, Trigger.dev, Warp and Zeabur. It also reported attacks on Chainstack, while WorkOS disclosed an attack on an internal application.

What data did the Chainstack attackers access?

Chainstack said an accessed database contained organization names, account-holder names, member email addresses and password hashes. SecurityWeek said node credentials might also have been exposed because attackers accessed systems storing them.

What information did the WorkOS attackers access?

WorkOS said hackers compromised an internal sales application and accessed business contact details and summaries of customer communications, including names, email addresses, phone numbers, job titles, company affiliations and customer account histories.

What is Artex in these attacks?

SecurityWeek reported that Chainstack and WorkOS disclosed cyberattacks that relied on Artex, an open-source AI penetration-testing framework, for automation.

How should affected customers respond?

Chainstack prompted customers to rotate node credentials. Fireworks instructed affected customers to replace revoked API keys, create new keys and update applications that used the revoked keys.

Sources

  1. “Black October” for AI Developer Platforms | Wiz Blog, WizPrimary
  2. Andromeda October 2026 Security Incident, AndromedaPrimary
  3. Fireworks October 2026 security incident, FireworksPrimary
  4. AI-Linked Breaches Mount as WorkOS, Northflank, Trigger.dev Disclose Hacks, SecurityWeek
How we checked this story
ClaimSourceStatus
SecurityWeek reported that at least a dozen AI developer platforms were hit over the past month.SecurityWeekAttributed
SecurityWeek said the hackers appeared to hunt for AI access tokens they could monetize and used AI agents to streamline attacks.SecurityWeekAttributed
SecurityWeek reported that the attacks did not appear related and that initial access varied across incidents.SecurityWeekAttributed
SecurityWeek said the attackers targeted Andromeda, Brev, CommandCode, Fireworks AI, Lightning AI, Northflank, OpenCode, Railway, Trigger.dev, Warp and Zeabur.SecurityWeekAttributed
The Cloud Security Alliance said attackers were reaching AI provider credentials through hosting, evaluation and developer-console intermediaries.Cloud Security Alliance, via SecurityWeekAttributed
Wiz said the rise was driven by LLM-jacking and LLM-broker infrastructure, which made stolen AI access tokens monetizable.Wiz, via SecurityWeekAttributed
Wiz observed multiple threat actors using the same compromised keys, suggesting a robust market for the credentials.Wiz, via SecurityWeekAttributed
SecurityWeek said chainstack was hit on October 26, and attackers accessed a development environment and parts of its production environment through a third-party component.SecurityWeekAttributed
Chainstack said an accessed database contained customer account information, including names, member email addresses and password hashes.Chainstack, via SecurityWeekAttributed
SecurityWeek said chainstack said node credentials might also have been exposed because attackers accessed systems storing them.SecurityWeekAttributed
WorkOS said hackers compromised an internal sales application on October 30 and accessed business contact details and customer-communication summaries.WorkOS, via SecurityWeekAttributed
WorkOS said the accessed information included names, email addresses, phone numbers, job titles and customer account histories.WorkOS, via SecurityWeekAttributed
SecurityWeek said multiple companies linked attacks attributed to Chinese APT Storm-0940 to Artex use.SecurityWeekAttributed
Wiz said two compromised enterprise credentials had identical first unauthorized activity within 30 minutes of reported credential access.WizAttributed
Andromeda said it identified unauthorized credential access on October 13, disabled compromised accounts that day and observed no further unauthorized activity.AndromedaConfirmed
Andromeda said the activity began as part of an industry-wide attack around October 9 and affected a limited number of customer accounts.AndromedaConfirmed
Fireworks said it identified unauthorized use of internal credentials on October 6 and that environment variables in limited customer accounts were accessed.FireworksConfirmed
wiz.io published its report on Oct 30, 2026.wiz.ioConfirmed
fireworks.ai published its report on Oct 10, 2026.fireworks.aiConfirmed

Could not verify

  • Whether the incidents were conducted by one group is not established
  • Whether the attackers exploited a specific vulnerability is not established
  • How many organizations and individuals were affected is not established
  • Whether Storm-0940 conducted the reported attacks is not established
See all in Data Leaks →
Explore with AI
ChatGPT
Claude
Perplexity
Grok