A threat actor has claimed the exfiltration of 25 million records and 14 GB of internal data from conversational commerce platform bitbybit Studio, allegedly exploiting an unpatched zero-day vulnerability in public-facing infrastructure.
Zero-day exploit targeting public API gateways or webhook endpoints
Backend access to PostgreSQL CRM, conversation transcripts, and audit logs
Unindexed query dumps generating 14 GB staging archive
25M records advertised across dark web cybercrime forums
If your team may be exposed to a similar threat, FemtoSec can help validate blast radius, prioritize remediation, and connect the issue to a practical security program.
August 11, 2026
A critical security flaw in the Vatican's Click to Pray application exposed the personal data of over 700,000 global users. This analysis highlights the technical details of the BOLA vulnerability, the registration hash leak, and practical steps to detect and prevent automated database harvesting.
August 9, 2026
An underground threat actor is actively selling two unauthenticated SQL injection zero-day vulnerabilities affecting widely used MyBB plugins with over 20,000 downloads. Learn the technical mechanics of the threat, the attack chain, and actionable detection strategies to secure your forum infrastructure.
July 15, 2026
An active exploit code for CVE-2026-50656, a high-severity local privilege escalation flaw in the Microsoft Malware Protection Engine, is being sold on dark web forums. Learn how the RoguePlanet exploit works and how to protect your enterprise Windows systems.
Exfiltration of 14 gigabytes of backend data containing 25 million structured records has highlighted critical exposure risks in cloud-integrated conversational platforms. Threat actor seraphims announced the compromise and underground sale of internal databases belonging to bitbybit Studio, an Indonesian artificial intelligence commerce platform facilitating merchant transactions across WhatsApp and Instagram. The exfiltrated dataset reportedly encompasses customer personally identifiable information, complete AI conversation logs, Meta and Facebook metadata, merchant contact broadcasts, and internal system security audit logs.
According to claims posted on illicit forums, initial access was gained through the exploitation of an unpatched zero-day vulnerability within the platform's internet-facing architecture. While third-party attribution and root-cause validation remain ongoing, the breach illustrates the severe systemic risks when conversational AI handlers and message ingestion endpoints interface directly with backend transactional databases without robust defense-in-depth controls.
Active exploitation status was confirmed through underground forum listings where threat actor seraphims provided samples of the exfiltrated dataset. The dataset includes sensitive PII such as full names, email addresses, phone numbers, delivery coordinates, Facebook platform metadata, and internal audit logs.
At present, no public proof-of-concept (PoC) script or technical exploit code has been shared openly in developer communities. However, the presence of internal system audit logs within the leaked sample indicates that the actor achieved elevated database query privileges, bypassing perimeter access controls. Security teams must treat conversational commerce webhooks and associated cloud microservices as high-value targets subject to automated probing.
The alleged compromise targets cloud-hosted conversational commerce services, specifically affecting production database clusters and webhooks exposed to third-party messaging integrations. While specific affected software versions have not been officially disclosed through vendor advisories, the threat actor claims the exfiltrated repository spans operational systems active through August 2026.
Organizations operating conversational commerce platforms face heightened exposure under specific architectural conditions:
Unsegmented Database Architecture: Backend PostgreSQL or NoSQL repositories storing message logs directly accessible from public-facing API gateways or microservices.
Insecure Webhook Handlers: Publicly reachable webhook listener endpoints that process incoming Meta Graph API or WhatsApp Business payloads without strict cryptographic signature validation or rate limiting.
Excessive Data Retention: Storing unredacted conversational histories, shipping addresses, customer contact details, and administrative logs in operational data stores indefinitely.
Broad Service Account Privileges: Application service roles possessing blanket read permissions across both transactional message streams and internal security audit logs.
While the threat actor claims that an unassigned zero-day flaw facilitated entry, typical technical root cause vectors for such platform compromises involve unauthenticated remote code execution (RCE), server-side request forgery (SSRF), or insecure direct object references (IDOR) on ingestion endpoints. Modern conversational commerce stacks aggregate public webhooks from Meta, e-commerce platforms like Shopify, and proprietary AI inference engines, presenting an expansive attack surface.
Security analysis indicates a structured four-stage intrusion path:
Initial Access: The adversary probes public-facing cloud endpoints, exploiting a zero-day flaw in API routing logic or webhook ingestion parsers to bypass gateway perimeter controls.
Privilege Abuse and Data Access: Leveraging compromised service credentials or SQL injection paths, the actor accesses core database clusters hosting customer relationship management tables, conversational transcripts, and security audit logs.
Bulk Query and Staging: Executing unindexed table dumps, the attacker stages approximately 14 GB of structured data into localized archives.
Exfiltration and Monetization: The archive containing 25 million records is transferred to external infrastructure and subsequently listed for sale on cybercrime marketplaces.
Continuous validation of exposed web services and API endpoints via Attack Surface Management is critical to discovering unmonitored routes before threat actors exploit them.
The scope of this compromise extends beyond standard identity theft, creating multifaceted operational, privacy, and regulatory risks across connected business ecosystems:
Conversational Data Exposure: AI chat transcripts frequently contain sensitive customer intents, proprietary business pricing negotiations, custom order details, and raw customer queries.
Targeted Phishing and Social Engineering: Attackers can leverage verified customer phone numbers, delivery addresses, and past order histories to conduct highly convincing smishing and spear-phishing campaigns.
Audit Log Weaponization: Exfiltrated system security audit logs provide threat actors with architectural blueprints of internal logging systems, user roles, IP ranges, and defensive blind spots.
Regulatory and Compliance Liabilities: Unlawful disclosure of personal identifiers triggers severe penalties under data protection regulations, including Indonesia's PDP Law, GDPR, and regional GCC data privacy frameworks.
Free exposure check
Dark Web Scanner
check dark web mentions, compromised account indicators, malware log signals, public breach exposure, and recent underground market activity for your domain.
Defending against zero-day exploits in complex API and conversational platforms requires proactive technical controls, rigorous network segmentation, and disciplined data sanitation practices.
Rotate All API Keys and Secrets: Immediately invalidate and re-issue Meta Business secrets, WhatsApp API tokens, cloud database master passwords, and administrative session tokens.
Enforce Ingress Verification: Validate HMAC signatures on all incoming webhook payloads and restrict inbound traffic to verified IP address ranges where supported.
Implement Micro-Segmentation: Place core databases inside private subnets unreachable from the public internet, requiring Zero Trust Network Access (ZTNA) or dedicated bastion proxies for administrative access.
Deploy Database Activity Monitoring (DAM): Implement alerting on abnormal query volumes, large-scale table exports, and unauthorized access to audit logging repositories.
Automate PII Redaction: Configure data pipelines to sanitize customer PII and payment indicators from conversational chat logs before persisting interactions to long-term storage.
Security operations teams should execute threat hunting queries against relational database activity to detect bulk data extraction patterns:
SELECT query_start, usename, client_addr, query, total_exec_time
FROM pg_stat_activity
WHERE state != 'idle'
AND (query ILIKE '%chat_logs%' OR query ILIKE '%customer_data%' OR query ILIKE '%audit_log%')
AND total_exec_time > 30000;Organizations seeking to uncover hidden flaws before malicious actors can exploit them should engage in regular Vulnerability Assessments and rigorous security code reviews across cloud-native platforms.