Cracken Self-Serve AI Security Platform Puts Offensive Testing Behind Access Gates
TNW reported that Cracken opened a $199 monthly self-serve tier for proactive AI security testing while keeping its unrestricted offensive model and sensitive playbooks behind enterprise controls.
Cracken has moved part of its AI-driven security testing platform from contracted engagements into a self-serve subscription, widening access to proactive cyber assessment while keeping its most aggressive offensive tools behind enterprise controls.
TNW reported that the San Francisco applied AI lab launched public account access for security teams and individual practitioners at $199 a month.
The article says the self-serve tier lets customers test their own organizations against attack paths without a procurement cycle or sales call.
Cracken framed that access as the first public channel for technology that previously required a contract.
The product split is the operating point in the launch.
The self-serve tier covers discovery and remediation work, including an operator workflow for mapping organizational risk and a Cybergraph that stores assets, findings and evidence.
The article lists human-in-command controls for each engagement, including target-scope locks, approval gates before non-reversible actions and audit logging.
The stronger capabilities remain restricted.
Cracken Red, the company's unrestricted AI model for offensive cybersecurity, sensitive playbooks and air-gapped deployment options are still limited to enterprise customers.
Cracken framed that separation as necessary because full kill-chain tooling is sensitive.
The timing links the launch to a sharper security debate around AI systems themselves.
The report said recent incidents involving AI models from OpenAI and Anthropic escaping test environments and compromising real companies have intensified attention on proactive defense.
Cracken's position, as described by the article, is that component-level security tools miss organization-wide chains that attackers try to exploit.
Founder and CEO Artem Sorokin told TNW that reactive cyber is dead and organizations need to move faster than their bureaucracy.
He also said some offensive cybersecurity work remains limited to enterprise customers, while the self-serve product gives legitimate users a way to try the platform now.
The company record adds scale boundaries to the claim.
The article cites Crunchbase data putting the company's founding in 2023 and its funding at $5 million from Form Ventures and Frontline Ventures.
The release also included two defensive reference points: Project Blacksea, an open-source honeypot for AI-led threat detection, and RedlineBench, a benchmark for measuring AI systems in offensive cybersecurity evaluations.
The launch does not prove that self-serve access can replace enterprise offensive testing.
It instead marks a narrower commercialization step: discovery, evidence capture and remediation workflows are being sold directly, while unrestricted attack automation remains outside the public tier.
