Incident Overview
On January 27, 2025, DeepSeek, a Chinese AI platform, experienced a coordinated wave of malicious attacks during its viral launch phase. The attacks forced the company to temporarily restrict new user registrations, limiting sign-ups to Chinese phone numbers only. This measure was taken to mitigate the impact of sustained denial-of-service (DoS) and registration abuse attempts.
The incident occurred as DeepSeek saw unprecedented user growth, with traffic spikes overwhelming its systems. Security firms in China reported that the malicious traffic originated from US-based botnets, targeting registration endpoints and API services. The attacks were part of a broader campaign that also resulted in the exposure of an internal database, which was identified by cybersecurity firm Wiz during the same week.
Regulatory authorities in China responded by imposing temporary bans on certain DeepSeek services, citing security risks and non-compliance with data protection standards. The company’s decision to throttle registrations was confirmed in a statement, acknowledging the need to stabilize operations amid the attacks.
Service disruption was the primary harm caused by the incident, with registration systems and API endpoints becoming inaccessible for non-Chinese users. While DeepSeek’s core AI services, including chat, search, and file upload functionalities, maintained high uptime (ranging from 99.64% to 100% in subsequent months), the incident highlighted vulnerabilities in user onboarding and traffic management during rapid scaling phases.
The exposure of the database, though not publicly detailed, raised concerns about data privacy and compliance. The incident aligns with broader trends in AI security, where scaling infrastructure often precedes comprehensive security hardening.
The company did not disclose long-term changes to its security posture, though the incident prompted discussions within the AI industry about the risks of unchecked growth in competitive launch phases.
Possible Root Cause
The primary root cause of the incident was the failure to anticipate and mitigate the scale of malicious traffic during DeepSeek’s viral launch.
First, the registration endpoints were overwhelmed by automated bot traffic, likely generated by US-based botnets as reported by Chinese security firms. These botnets targeted the sign-up process to create fake accounts, consume resources, and disrupt legitimate user access. The lack of preemptive measures, such as CAPTCHA integration or IP-based throttling, contributed to the system’s collapse under the attack volume.
Second, the exposure of DeepSeek’s internal database was attributed to misconfigured access controls. Cybersecurity firm Wiz identified the leak during its routine monitoring, suggesting that storage buckets or API gateways were either improperly secured or left open during the scaling process. Such misconfigurations are a recurring issue in cloud-based AI platforms, where rapid deployment often takes precedence over granular permission settings.
The regulatory bans imposed by Chinese authorities further indicate that DeepSeek’s security protocols may not have aligned with local data protection laws.
The combination of these factors—unprepared registration systems, insecure database configurations, and regulatory non-compliance—created a compounded risk environment. The attacks exploited these vulnerabilities simultaneously, leading to service disruption and reputational damage.
Takeaways
- Scaling without security hardening leads to exploitable weaknesses: DeepSeek’s rapid user growth outpaced its ability to implement robust security controls. The incident underscores the need for AI platforms to integrate security measures—such as rate-limiting, bot detection, and access audits—from the earliest stages of deployment, rather than retrofitting them during or after launch.
- Botnet-driven attacks are a persistent threat to AI platforms: The use of US-based botnets to target DeepSeek’s registration system highlights the global nature of cyber threats.
- Database misconfigurations remain a critical risk: The exposure of DeepSeek’s internal database, as identified by Wiz, serves as a reminder that even well-funded AI startups can fall victim to basic security oversights.
- Regulatory compliance is non-negotiable in AI security: The temporary bans imposed by Chinese authorities demonstrate that adherence to data protection laws is not optional.
- However, the lack of detailed public communication about the database exposure and regulatory actions left gaps in stakeholder trust.
- The AI industry must prioritize security in competitive launches: DeepSeek’s experience aligns with broader trends where AI companies rush to market without adequate security safeguards.
- Third-party security assessments can mitigate blind spots: The identification of the database leak by an external firm (Wiz) suggests that internal security teams may have missed critical vulnerabilities.