Anthropic faces significant security and competitive hurdles as it pushes Claude 3.5 Opus into the enterprise market. The company discovered 24,000 fraudulent accounts linked to Chinese AI labs DeepSeek, Moonshot AI, and MiniMax. These organizations allegedly used the accounts to generate 16 million interactions to perform illicit distillation. This process involves training weaker models using outputs from stronger models to mimic their capabilities. Anthropic warns that such distilled models lack the technical and ethical safeguards found in the parent product. These unprotected models increase the risk of offensive cyber operations, disinformation campaigns, and mass surveillance.
The company also deals with a reputation for opacity. Critics note that Anthropic, like OpenAI, has histories of training models on large datasets that sometimes include copyrighted material or personal data. Public figures on social media demand that Anthropic release full forensic reports for independent review. While Anthropic maintains a safety-focused mission, the competitive pressure from OpenAI forces difficult decisions. OpenAI is gaining traction with its Astra model, which threatens Anthropic’s strength in the enterprise sector. This competition forces Anthropic to choose between maintaining its safety-first release schedule and matching the velocity of rivals.
OpenAI rivalry and the coding market
OpenAI is mounting a direct challenge to Anthropic’s dominance in the enterprise coding space. Anthropic currently holds a 42 to 54 percent market share in enterprise coding, while OpenAI holds 21 percent. Coding represents 51 percent of all generative AI enterprise usage. Anthropic’s Claude Code, a terminal tool, generates $2.5 billion in annualized revenue. OpenAI’s Astra model is gaining traction among large business customers who prioritize recurring usage and long-term contracts.
The pricing battle between the two companies remains intense. OpenAI’s GPT-5.6 Sol is currently 20 percent cheaper than Claude Opus 5 for short-context work.
| Feature | Claude Opus 5.5 | GPT-5.6 Sol |
|---|---|---|
| Input Price (per 1M tokens) | $4 | $5 |
| Output Price (per 1M tokens) | $20 | $30 |
| Context Window | 1,000,000 tokens | 1,050,000 tokens |
| Max Output Tokens | 128,000 | 128,000 |
OpenAI maintains a price advantage for high-volume, low-cost workloads with its Luna model. GPT-6 Luna costs $0.20 per million input tokens and $1.20 per million output tokens. In comparison, Claude Haiku 4.5 costs $1 per million input tokens and $5 per million output tokens. OpenAI’s Luna is approximately 78 percent cheaper than Haiku 4.5 for a workload of 1,000 requests with 2,000 input and 500 output tokens.
Security vulnerabilities and deployment hurdles
Enterprise users face unique risks when integrating Claude into their existing workflows. Employees often submit proprietary or regulated information into prompts, which creates data exposure and compliance risks. Shadow AI usage occurs when teams adopt Claude without central approval, leaving security teams with no visibility into usage patterns. Organizations must manage long-lived API keys and ensure access ties to role-based policies to prevent data leakage.
Anthropic faces a different kind of scrutiny regarding government contracts. In early 2026, the Pentagon demanded that Anthropic remove restrictions preventing Claude from being used for mass domestic surveillance or autonomous weapons systems. Anthropic refused the request. The Trump administration subsequently designated the company a supply chain risk. This designation follows Anthropic’s decision to hold its position past the February 27 deadline.
The company also faces reliability issues. In April 2026, users reported spikes in errors on Down Detector. Anthropic confirmed the issue impacted logins, voice mode, and standard chat functions. Even though the company applied a fix, these outages disrupt businesses that rely on Claude for daily operations. Will Anthropic’s refusal to compromise on safety restrictions eventually lead to the loss of major government contracts?
