Why ?

Detailed Analysis

The Reddit post captures a strain of grassroots skepticism toward Anthropic and the broader AI industry, prompted by the poster's encounter with several linked stories: Anthropic's own disclosure about investigating cybersecurity incidents involving its models, a BBC report, and CNBC coverage of security issues touching OpenAI and Hugging Face. Rather than analyzing any single incident in depth, the author uses these links as a jumping-off point for a broader complaint — that AI companies market their tools as transformative and safe while simultaneously gatekeeping access, restricting who can use their most capable systems, and facing little accountability compared to how aggressively they or law enforcement pursue independent actors who misuse or reverse-engineer software. The post reflects frustration with what the author perceives as a double standard: large AI labs conducting or enabling activities (automation, security research, model probing) that would be treated as suspicious or malicious if performed by individuals or smaller entities without institutional backing.

The specific trigger here — Anthropic's "Investigating Incidents Involving Cybersecurity" disclosure — is part of a genuine and increasingly important trend in AI safety reporting. Anthropic has been relatively transparent, compared to some peers, about publishing findings when its models are misused for cyberattacks, fraud, or other harmful activity, including detailed writeups of specific incidents where threat actors attempted to leverage Claude for hacking, social engineering, or malware development. This kind of disclosure is meant to demonstrate responsible stewardship and to inform the security community, but it can also read, to outside observers, as evidence that these systems are powerful enough to be dangerous — reinforcing the poster's sense that capability is being tightly held by a few well-resourced companies while access for others is throttled through KYC-style verification, usage policies, and account bans for suspected malicious intent.

The complaint about restricted access is a real and recurring friction point in the AI industry. Frontier labs increasingly gate advanced capabilities (higher rate limits, agentic tool use, coding assistants with system access, jailbreak-resistant models) behind business verification, enterprise contracts, or usage-policy enforcement that disproportionately affects hobbyists, security researchers, and users in the reverse-engineering or automation communities — some of whom report account suspensions with little explanation or recourse. This creates a perception of a two-tier system: well-funded organizations can access and shape frontier AI capabilities, while individual developers face suspicion by default. Anthropic, OpenAI, and others frame this as necessary risk mitigation given the dual-use nature of powerful models, but critics see it as a mechanism that concentrates power and monetization opportunity among incumbents while publicly emphasizing safety and democratization.

Broadly, this post is symptomatic of a widening trust gap between AI companies and segments of the technical public. As labs publish more safety and incident-response content — intended to build credibility — some of that same material fuels suspicion that the technology's real capabilities and risks are being managed opaquely, with commercial incentives dressed in safety language. This tension sits alongside ongoing industry-wide debates over responsible disclosure, model access tiers, and the enforcement of acceptable-use policies, and it foreshadows continued friction as AI labs try to balance broad accessibility, commercial growth, and risk mitigation, especially as capabilities in areas like autonomous cyber operations and agentic coding continue to advance and draw regulatory and public scrutiny.