Why do hackers impersonate AI bots?

Security systems and administrators have grown accustomed to trusting traffic from well-known AI services: DeepSeek, ChatGPT, Claude. By spoofing the User-Agent header, attackers reduce the chance of triggering WAF rules — some attacks slip through undetected. This exploits organisational trust, not a software vulnerability.

What is DNS Rebinding and why is it dangerous?

DNS Rebinding is a technique where an attacker manipulates DNS responses so that the victim's browser treats a malicious site as an internal network resource. This lets the attacker reach devices behind NAT — home routers, IP cameras, internal corporate services — bypassing conventional perimeter defences. This type of attack accounts for 53% of the AI-bot-masked requests found by Solar WAF.

How can an ordinary user protect themselves against these attacks?

Individual users have no direct defence against attacks on corporate websites — that is the responsibility of service owners. However, you can limit the impact: use a password manager with unique passwords per service, enable two-factor authentication, and keep your browser and OS updated. LiMP VPN encrypts your traffic and hides your IP address, reducing the risk of third-party surveillance of your online activity.

← All news

September 20, 2026

Hackers Mask Web Attacks as ChatGPT and DeepSeek Bots

Emily ParkerContent & SEO Specialist

Hackers Mask Web Attacks as ChatGPT and DeepSeek Bots

In short: Since August 12, 2026, attackers have been spoofing HTTP User-Agent headers to make malicious requests look like traffic from ChatGPT, DeepSeek, Claude, Grok and Perplexity. Solar WAF analysts at cybersecurity firm Solar uncovered the campaign: 53% of disguised requests are DNS Rebinding attacks targeting internal networks, while 12% aim directly at data exfiltration. The technique exploits institutional trust, not a software bug.

How attackers impersonate AI bots

Every HTTP request carries a User-Agent header — a string that tells the server what is connecting: a browser, a search crawler, or an AI service. Security teams and automated defences have grown accustomed to seeing DeepSeekBot, ChatGPT-User or Claude-User in logs, and often treat such traffic as low-risk.

Starting August 12, 2026, Solar WAF began detecting malicious requests carrying spoofed strings: DeepSeekBot, ChatGPT-User, Perplexity-User, Claude-User and GrokBot. A familiar AI name raises less suspicion and increases the chance of slipping past rule-based filters unchecked. This pattern was entirely absent from Solar WAF telemetry in June and July 2026 — it emerged in August and spread quickly. Stay up to date on evolving threats in the LiMP VPN security blog.

Four attack categories hiding behind the AI mask

Solar WAF's analysis breaks down the disguised malicious activity into four categories:

53% — DNS Rebinding. By manipulating DNS records, attackers bypass the browser's Same-Origin Policy and reach internal network resources — routers, IP cameras, corporate services not directly exposed to the internet.
12% — Data exfiltration attempts. Requests designed to extract confidential information from web applications: source code fragments, configuration files, user records.
4% — Path Traversal. Manipulating file paths to escape the permitted directory and read arbitrary files from the server.
31% — Other attacks, including SQL injection and other classic web-application exploit techniques.

Why this technique is particularly dangerous for your data

The User-Agent spoofing attack exploits organisational trust, not a software vulnerability. If a WAF or security team has configured relaxed rules for traffic from known AI crawlers, the attacker has already bypassed half the defence without touching a single file.

DNS Rebinding can silently allow attackers to control devices on your home or office network — routers, smart TVs, IP cameras — directly through the browser of any user who visits a compromised page. Data exfiltration from corporate web applications means your personal details — phone number, address, order history, ID document data — can end up in the wrong hands. An encrypted tunnel via LiMP VPN hides your real IP and traffic from third-party observers, adding a layer of privacy even when a third-party service's perimeter has been breached.

What security experts recommend

Alexey Pashkov, head of WAF at Solar, identifies three key protective measures:

Multi-layer defence. A single WAF does not cover every attack vector — the right combination is firewall, WAF, anti-bot system, and anomaly monitoring working together.
Machine-learning anti-bot systems. Classic User-Agent rules are trivially bypassed by header spoofing. ML-based systems analyse request behaviour — timing, navigation chains, TLS fingerprints — which is far harder to convincingly fake.
100+ request parameter analysis. User-Agent is just one signal. A genuine AI crawler differs from an attacker across dozens of other attributes: request frequency, URL patterns, connection signatures.

For individual users, the practical takeaway is straightforward: keep software and browsers updated, use unique passwords per service, and enable two-factor authentication. Data leaks from corporate websites are largely outside your control — but a LiMP VPN subscription reduces the risk of third-party traffic surveillance on your end.

Sources

← All news