High-Performance Model Launch and Ethics Discourse Simultaneously... Anthropic Builds Momentum Ahead of IPO (Comprehensive)

Claude Fable 5 Released with Safety Features

Active Moves Ahead of IPO, with Ethical Issues Also Addressed

Global artificial intelligence (AI) company Anthropic has released to the public its high-performance AI model, which previously raised security concerns under the name 'Mythosval.' With its competitor OpenAI preparing for an initial public offering (IPO), Anthropic has also launched highly capable AI models in an effort to build momentum ahead of its own IPO.

On June 9 (local time), Anthropic announced the launch of its high-performance Mythos-class AI model, 'Claude Fable 5,' and its security-focused model, 'Claude Mythos 5.' Due to concerns over potential misuse in cybersecurity, Anthropic had previously made Mythos available only to select companies and institutions. The newly released Fable 5 model incorporates safeguards for high-risk areas such as cybersecurity and biological and chemical research. Anthropic explained, "We have conservatively adjusted safety controls to ensure the model is released both safely and quickly," adding, "There may be cases where even benign requests are declined, but these account for less than 5% of all cases." As before, Mythos 5 will be provided exclusively to companies and institutions participating in the global security consortium, 'Project Glasswing.'

Anthropic emphasized that the performance of its newly released AI models has improved significantly compared to the 'Mythos Preview' introduced in April. Mythos 5 scored 78% in the 'ExploitBench' test, which measures cybersecurity-related abilities. This surpasses not only Mythos Preview's 69% but also OpenAI's GPT-5.5 (34%) and Anthropic's own Opus 4.8 (40%). In the 'Humanity's Last Exam' (HLE), which indicates doctoral-level intelligence, Mythos 5 achieved a score of 59%, outperforming Mythos Preview's 56.8%.

Although the Fable 5 model, equipped with safety mechanisms, does not match Mythos 5 in raw performance, it has seen improvements in programming and analytical capabilities. In the 'SWE-Bench Pro,' which evaluates AI coding skills, Fable 5 recorded a success rate of 80.3%. In contrast, Opus 4.8 had a success rate of 69.2%, and GPT-5.5 achieved only 58.6%. Anthropic stated, "Fable 5 shows outstanding performance in complex analytical tasks," and added, "There have been significant improvements in document-based reasoning, chart and table interpretation, and problem-solving abilities."

Anthropic is seen to be accelerating its growth trajectory by releasing high-performance AI models ahead of its IPO. On June 1, Anthropic submitted a confidential IPO filing to the U.S. Securities and Exchange Commission (SEC). As the number of users has surged recently, Anthropic appears to be gaining ground against OpenAI. In the recently completed Series H investment round, Anthropic was valued at $965 billion (approximately 1,465 trillion won), surpassing OpenAI's $852 billion. Regarding Fable 5, Anthropic stated, "Demand is expected to be very high," and announced that from June 23, a separate fee will be required to use Fable 5.

At the same time as its rapid growth, Anthropic is also actively addressing ethical issues related to AI. Following a conflict with the U.S. Department of Defense over the use of AI in warfare, the company's user base has skyrocketed, and it is seeking to carry this positive momentum into its IPO. On June 4, Anthropic stated on its blog, "AI may soon reach a stage where it can improve itself without human intervention," and argued that "the pace of cutting-edge AI development should be slowed to allow social systems and AI research to keep up with technological advances."

This content was produced with the assistance of AI translation services.