PORTALBERITA.CO.ID - Anthropic has officially made its new artificial intelligence Fable 5 available to the wider public, according to The Guardian. This significant release is the first general access version stemming from the company's high-capability Mythos class, initially unveiled in April.
The model boasts powerful functionalities, including effective software code writing and debugging capabilities, along with the ability to address complex research inquiries and perform image analysis. However, its deployment has been carefully managed due to inherent security considerations.
Prior to this public launch, Fable 5 was restricted for several months, accessible only to a select group of partner institutions involved in testing and validation. This controlled rollout aimed to address potential cybersecurity vulnerabilities associated with the advanced technology.
In a parallel structure, Anthropic continues to provide an unrestricted iteration, known as Claude Mythos 5, exclusively to established enterprise clients already utilizing the Mythos family. This privileged access group notably includes cybersecurity partners enrolled in the ongoing Project Glasswing initiative.
The Project Glasswing cohort, which assesses the model's security implications, expanded significantly in early June, now encompassing roughly 200 organizations across more than 15 nations, with further scaling anticipated. The core reason for these infrastructure restrictions relates to the model’s demonstrated capacity to rapidly identify weaknesses within critical systems, such as power grids and banking platforms.
When Project Glasswing first commenced, some detractors suggested the startup was exaggerating potential dangers to garner attention. Nevertheless, organizations that subsequently tested the Mythos class have since validated its advanced operational capacities, lending credence to Anthropic's security concerns.
The United States government, which had previously engaged in a legal disagreement with Anthropic over security matters, has also conducted its own evaluation of the model. Following these assessments, the White House formalized an agreement requiring leading AI companies to test their most potent models internally before any public release.
Anthropic has implemented automatic routing protocols to ensure user safety and control proprietary information. "Most queries submitted to Fable 5 regarding cybersecurity, biology, or chemistry will instead be automatically routed to its lower-tier model, Opus 4.8," the company stated.
This fallback mechanism also addresses attempts to exploit the technology for illicit purposes. "Queries originating from these types of activities will similarly fall back to the less powerful model," Anthropic confirmed, aiming to prevent the extraction of proprietary technology for training competing models in authoritarian territories.
