An offer of 90 percent off the official price to access Claude, the powerful AI model from Anthropic, proved irresistible to roughly 900 people. That temptation now appears to have been a costly mistake. The scheme, which exploited the growing demand for premium AI tools, managed to deceive hundreds of users before it was identified. This incident is not merely a cautionary tale about a single discount scam; it represents a significant escalation in the tactics used by cybercriminals to capitalize on the artificial intelligence boom.
How the 90 Percent Off Claude Scam Worked
At the heart of the deception was an offer that seemed too good to pass up. Victims were presented with what appeared to be a legitimate pathway to access Claude, Anthropic’s flagship large language model, at a fraction of its standard subscription cost. The scam promised a 90 percent discount, a reduction so dramatic that it should have raised immediate red flags. Yet, for approximately 900 users, the allure of gaining premium AI capabilities at a near-negligible price overrode their skepticism.
The mechanics of the scam are emblematic of a broader category of phishing and credential harvesting attacks. Users who took the bait were likely directed to a fraudulent sign-up page that mimicked Anthropic’s official interface. By entering their payment details and, in many cases, their API keys or login credentials, victims inadvertently handed over sensitive information directly to the scammers. Instead of gaining discounted access to Claude, they lost control of their accounts and potentially exposed their data to unauthorized parties.
What is a Claude API key scam? A Claude API key scam is a fraudulent scheme where criminals trick users into providing their Anthropic API credentials by offering fake discounts, free trials, or exclusive access to Claude models. Once obtained, these keys can be used to run expensive inference queries at the victim’s expense, or sold on dark web markets to other malicious actors.
The financial damage extends beyond the initial fee paid for the fraudulent discount. Victims may face unauthorized charges running into hundreds or thousands of dollars if their API keys were harvested and used for bulk inference tasks. The scam effectively weaponized the very trust that Anthropic has built around its brand, demonstrating that even well-educated technology users remain vulnerable to carefully crafted social engineering campaigns.
The Scale of the Deception: Why 900 Victims Matters
In the context of online fraud, 900 compromised accounts might seem like a modest number. Major data breaches often involve millions of records. However, the significance of this incident lies not in the raw count of victims but in what it reveals about the evolving threat landscape surrounding generative AI.
Nine hundred users represent a highly targeted group. Unlike broad, spray-and-pray phishing campaigns that blast millions of emails in the hope of catching a few clicks, this scam appears to have been calibrated specifically at individuals already seeking access to Claude. The perpetrators demonstrated an understanding of their victims’ intent, timing the offer to intercept users at the moment of purchase intent. This level of precision suggests that the attackers conducted reconnaissance on forums, social media groups, and other communities where developers and AI enthusiasts discuss tools like Claude.
The financial implications are also more severe than they appear. If each victim paid, for example, a nominal fee of $10 to $20 for the supposed discount access, the direct take for the scammers could be between $9,000 and $18,000. However, the indirect damage from stolen API credentials could multiply that figure many times over. Attackers who harvest API keys can use them to query Claude at full price, racking up charges on the victim’s account before the fraud is detected. In some cases, stolen API keys are bundled and resold, enabling further abuse by additional parties.
The Broader Pattern of AI-Themed Scams
This incident fits into a larger and troubling pattern. As large language models like Claude, GPT-4, and Gemini have moved from experimental curiosities to essential productivity tools, they have become prime targets for fraud. Scammers are adapting classic online cons to the AI context, creating fake subscription portals, counterfeit API key generators, and fraudulent plugin marketplaces.
The Claude discount scam shares DNA with previous phishing operations that targeted ChatGPT Plus subscribers. In those cases, users received emails offering discounted access to GPT-4, only to have their credentials stolen. The playbook is identical: identify a popular premium service, craft a compelling economic incentive, and deploy a convincing copy of the official interface. The success of these campaigns depends on the fact that many users genuinely want lower prices, creating a psychological vulnerability that scammers exploit ruthlessly.
Anthropic’s Claude is an especially attractive target because of its reputation for safety and its strong performance in coding and analytical tasks. Developers and researchers who rely on Claude for their work are willing to pay for access, making them a demographic with both the technical skill to understand API usage and the financial incentive to seek deals. The scammers understood this intersection perfectly.
Why Users Fell for the Claude Discount Offer
Understanding the victim psychology is essential to preventing future incidents. Why would approximately 900 people, many of whom are likely technically literate, fall for what in retrospect appears to be an obvious scam? The answer lies in a combination of cognitive biases and environmental factors.
The Scarcity and Urgency Trap
Fake discount offers almost always create a false sense of urgency. Messages promoting the 90 percent off Claude deal likely included phrases like “limited time offer,” “only 50 spots remaining,” or “exclusive early access.” This pressure short-circuits rational decision-making. The fear of missing out overrides the instinct to verify the offer’s legitimacy. When a deal appears to be expiring, the brain prioritizes speed over verification.
Brand Trust Exploitation
Anthropic has cultivated an image of integrity and safety. The company’s emphasis on responsible AI development has earned it considerable goodwill in the tech community. Scammers exploited this trust by crafting communications and landing pages that mirrored Anthropic’s official branding. Victims did not suspect fraud because they associated the visual language of the scam with a company they trusted. This is a classic brand impersonation attack, but one that gains extra potency when the brand in question is associated with cutting-edge technology and ethical values.
The Familiarity of Subscription Fatigue
Modern technology users are inundated with subscription services. The typical knowledge worker may pay for ChatGPT, Claude, GitHub Copilot, Slack Premium, Zoom Pro, and a dozen other tools. Subscription fatigue creates a genuine desire for discounts and bundles. When an offer promises to cut one of those recurring costs by 90 percent, it resonates with a real pain point. Scammers are increasingly leveraging subscription fatigue as a vector, because it makes their offers feel less like a trap and more like a solution.
How to Verify a Legitimate Claude Discount or Offer
Distinguishing between a real promotion and a fraudulent scheme requires a systematic approach. The following steps can help users verify the authenticity of any Claude-related offer before parting with their money or credentials.
First, always navigate to the official Anthropic website directly by typing the URL into the browser. Never click links from emails, social media messages, or third-party forums. Legitimate promotions from Anthropic will be advertised on their official domain, and any discrepancy in the URL is a major red flag. Second, check for official announcements. Anthropic typically communicates pricing changes and promotions through their official blog, social media accounts, and press releases. If a discount is not mentioned in these channels, it is almost certainly fraudulent. Third, inspect payment gateways carefully. Legitimate transactions for Claude subscriptions will be processed through recognized payment providers and will redirect to secure checkout pages. Requests for cryptocurrency payments, wire transfers, or gift cards are definitive indicators of a scam.
Fourth, users should never share API keys outside of the official Anthropic console. API keys are the digital keys to a user’s account and usage quota. No legitimate discount offer requires an API key to apply a coupon or promotional code. Fifth, look for contact and support information. Scam sites often omit valid contact details or provide only a generic email address. Checking the site’s registration date and WHOIS information can also reveal whether the domain was created very recently, a common tactic for short-lived phishing operations.
The Technical Infrastructure Behind AI Scams
The sophistication of the Claude discount scam points to a well-organized technical operation. These attacks do not happen by accident. They require investment in infrastructure, including domain registration, website hosting, email delivery systems, and credential harvesting backends.
Modern phishing kits for AI services have become remarkably advanced. They can replicate the entire user interface of Anthropic’s dashboard, including dynamic elements like usage statistics and model selection menus. Some kits even include real-time validation of submitted credentials, checking them against the actual Anthropic API to confirm they work before storing them. This level of polish makes the fake site indistinguishable from the real one to a casual observer.
The attackers likely used domain names that closely resemble Anthropic’s official domains. Typosquatting, where a domain like anathropic-claude.com or claude-discount.net is registered, is a common technique. More sophisticated operations use homograph attacks, substituting characters from other alphabets that look identical to Latin letters. A Cyrillic “а” instead of a Latin “a” in a URL is invisible to the human eye but directs the browser to an entirely different server.
Email delivery for these campaigns is often routed through compromised WordPress sites or hijacked SMTP servers, making it harder for spam filters to block the messages. The entire infrastructure is designed for short lifespan. Domains are used for a few days or weeks before being abandoned, forcing authorities and security researchers to play a constant game of whack-a-mole.
The Role of Social Engineering in the Claude Scam
Technical infrastructure alone does not explain the success of this scam. Social engineering was the critical ingredient. The perpetrators did not just build a fake website; they built a narrative. The story of a 90 percent discount on Claude fits seamlessly into the broader discourse around AI pricing. Many users genuinely believe that AI services are overpriced and that discounts are inevitable as competition intensifies. The scam tapped into that belief system.
Social engineering also extends to the channels used to promote the offer. Forums, Discord servers, Telegram groups, and Reddit communities dedicated to AI development are fertile ground for this type of attack. Scammers or their accomplices post messages praising the deal, often using multiple fake accounts to create an illusion of social proof. A prospective victim sees a thread where several users claim to have successfully redeemed the discount, which dramatically increases the perceived legitimacy of the offer. This manufactured consensus is a powerful psychological lever.
The scam also likely used a technique called “authority hijacking.” If the promotion appeared to be endorsed by a respected community member or was shared in a channel dedicated to legitimate AI news, users would lower their guard. In some cases, scammers compromise the accounts of real community members and use those accounts to spread the fraud, adding an additional layer of deception.
What Anthropic and Other AI Companies Can Learn
This incident carries clear lessons for Anthropic and the broader AI industry. Companies that build popular AI tools have a responsibility not just to secure their own infrastructure but to help users navigate a threat landscape that increasingly targets their brands.
First, proactive communication is essential. When a scam like this is identified, companies should issue immediate public warnings through every available channel, including email, blog posts, social media, and even in-app notifications. The speed of the response directly correlates with the potential to limit damage. Second, user education must be continuous. AI users need to understand that API keys are as sensitive as credit card numbers and should never be shared with third parties. Companies should consider adding mandatory security briefings or tooltips that explain key safety practices.
Third, technical countermeasures can help. Anthropic could implement domain monitoring services that detect lookalike domains and initiate takedown requests automatically. Browser extensions that verify the authenticity of login pages could be developed or promoted. Multi-factor authentication should be mandatory for all API access, adding a layer of protection even if credentials are stolen.
Fourth, pricing transparency works as a security measure. If Anthropic clearly communicates that it does not offer discounts through third-party channels or unsolicited emails, users are less likely to be swayed by fraudulent offers. Establishing a verified partner program could also help users identify legitimate resellers or promotional partners if any exist.
The Economic Model of AI Subscription Fraud
To understand why scams like the Claude discount offer persist, it is necessary to examine the economic incentives behind them. AI subscription fraud is a business, and like any business, it operates on margins, volume, and customer acquisition cost.
The primary asset sought by these scammers is not the upfront payment for the fake discount, but the API credentials themselves. Stolen Claude API keys have a thriving secondary market. They are used for a variety of purposes: running inference jobs without paying for compute, accessing Claude for banned use cases that Anthropic would block, or reselling access to users who cannot obtain it legitimately due to geographic restrictions. A single compromised API key can yield significant value before it is revoked, especially if the victim has a high-usage plan.
The cost of running a phishing campaign is relatively low. A domain name costs a few dollars. Phishing kits can be purchased or rented on dark web forums for minimal sums. Email lists of potential victims can be scraped from AI discussion communities. With a potential return of thousands of dollars per campaign, the economics strongly favor the attackers. This asymmetry between low cost and high reward ensures that AI-themed phishing will remain a persistent threat.
Law enforcement faces significant challenges in combating these operations. The cross-border nature of cybercrime, combined with the use of anonymous payment systems and encrypted communications, makes attribution and prosecution difficult. The scammers behind the Claude discount campaign may never be identified, and even if they are, jurisdictional hurdles could prevent legal action.
Practical Steps for Users Who May Have Been Affected
Anyone who suspects they may have fallen victim to this or a similar scam should take immediate action. The first step is to change all passwords associated with the compromised account, ensuring that the new password is strong and unique. If the same password was used across multiple services, it must be changed everywhere immediately.
The second step is to revoke any API keys that may have been shared with the fraudulent site. This can be done through the official Anthropic console. Once old keys are revoked, new keys should be generated and stored securely. Victims should also review their account activity for unauthorized usage. Anthropic provides logs of API calls and billing details. Any unexpected activity should be reported to Anthropic’s support team.
Third, victims should monitor their financial accounts for unauthorized charges. If the scam collected payment card details, those cards should be frozen or replaced. Reporting the fraud to the card issuer may also help in disputing any charges that were made without authorization.
Fourth, filing a report with relevant cybercrime authorities can help build a case against the perpetrators. In the United States, the FBI’s Internet Crime Complaint Center (IC3) accepts reports of online fraud. Similar agencies exist in other countries. While individual reports may not lead to immediate action, aggregated data helps law enforcement identify patterns and prioritize investigations.
Finally, sharing information about the scam with the community can help prevent others from falling victim. Posting details about the fraudulent domain names, email addresses, and messaging tactics on security forums or social media raises awareness and shortens the operational lifespan of the scam infrastructure.
The Claude AI discount scam that duped 900 users is a stark reminder that the generative AI era has a dark underbelly. As the technology becomes more embedded in daily work and life, the incentives for criminals to target its users will only grow stronger. The technical community must respond with equal sophistication, combining better security practices, more robust platform protections, and a culture of healthy skepticism. The best defense against a 90 percent discount that feels too good to be true is the simple, timeless rule that it probably is.
