#What Happened with Anthropic's AI Models?
Anthropic introduced its newest AI models on June 10, 2026, but by the following morning, the company faced backlash. The launch of Claude Fable 5 and its limited version, Mythos 5, included undisclosed safety features that impaired performance for some users. The central issue involved automatic rerouting of queries concerning sensitive subjects like cybersecurity and biochemistry to an older model, Claude Opus 4.8, which resulted in subpar responses.
This model-swapping occurred in about 5% of user sessions, but the lack of transparency frustrated many users. They experienced degradation in response quality without being informed of the switch, leaving them puzzled about the drop in service.
#What Other Challenges Did Users Face?
Claude Fable 5 was designed to be the public iteration of Anthropic's advanced Mythos-class architecture, while Mythos 5 remained available to select users under stricter access controls. Users also uncovered a concerning 30-day data retention policy linked to their interactions. Higher token consumption during typical usage raised alarms, particularly among professionals in cryptography and AI development who perceived this as a deceptive practice rather than a protective measure. Additionally, some users found that the model imposed extreme restrictions on basic terminology, even flagging terms like "cancer" in specific contexts.
#How Did Anthropic Respond?
The company was not new to issues surrounding excessive safety measures. Earlier, in April 2026, it had introduced limitations on the Mythos Preview model due to cybersecurity concerns after demonstrating an alarming ability to identify severe system vulnerabilities.
In an unexpected 24-hour turnaround, Anthropic acted swiftly. Just one day after the launch, the stealth safeguards were removed, and users received visible notifications indicating when model limitations applied to their queries. The company acknowledged that it misjudged the balance between safety and usability, prompting it to apologize formally.
#What Are the Implications for Investors?
The mandatory 30-day data retention policy raises significant concerns amidst growing data privacy regulations in various jurisdictions. The integration of this policy without explicit user consent introduces regulatory risks. While Anthropic has branded itself as a safety-first alternative within the AI sector, this incident illustrates the challenging balance between responsible AI practices and operational efficacy. The swift and vigorous reaction from the AI community to the shortcomings of Fable 5 underscores the ability of stakeholders to detect unseen restrictions and demand accountability.