Where did the "Cancer Secret" idea come from?
The conception about "hidden medical secrets" or "hiding the existence of cancer" comes from overly aggressive platform safety filters triggering false positives.
When Anthropic and other tech companies deploy safety classifiers to prevent people from using powerful AI models for illegal or dangerous tasks (such as bio-weapon design, unauthorized medical advice, or hacking), the automated systems act like a stupid blunt instrument.
• If a regular user prompts the model using highly regulated keywords—such as complex medical terminology mixed with technical code—the AI safety classifier often panics.
• The system triggers an automated platform-wide lockdown or censors the data to avoid high-risk liability.
• When the AI suddenly refuses to answer or states it cannot discuss the topic, some users interpret this automated compliance block as a conspiracy to hide medical "secrets" from the general public. Some idiotic users think that this stupid “superintelligence” has some good unknown reason to discredit American AI this way.
“NEW YORK. Anthropic, a company specializing in artificial intelligence, is taking a new approach with a product that has recently been the subject of intense discussion, making it available to the general public. The product in question is the "Mythos" AI model. Launched in April, it possesses—according to Anthropic—advanced capabilities for identifying and exploiting software vulnerabilities. This makes it valuable not only to software vendors but also to hackers. Due to this security risk, Anthropic initially restricted access to Mythos to a select group of users, including tech giants like Microsoft, Google, and Nvidia.
Now, Anthropic has announced the release of a "Mythos-class" model for the general public, albeit one equipped with built-in safeguards.
Named "Claude Fable 5," the model is programmed to block queries regarding specific sensitive topics and redirect them to "Claude Opus 4.8," another Anthropic model.
Specifically, these topics include cybersecurity and biology. The aim is, among other things, to prevent the model from being misused for hacking or the development of biological weapons.
Anthropic states that these safeguards are currently set to a "conservative" level. They are described as "stricter than ideal" and may filter out even "harmless queries."
Some users have noted that Fable 5 also blocks and redirects queries related to diseases such as cancer.
Anthropic has promised to address this and reduce "false positives" as quickly as possible. Overall, however, early tests indicated that the Fable model answers more than 95 percent of all queries without redirecting them. In these instances, Fable 5 functions essentially the same way as "Mythos 5"—the newly unveiled version of the Mythos series that remains restricted to a limited user group.
Access to both Fable 5 and to the "Mythos 5" models requires payment.
Anthropic describes "Fable 5" as its most powerful model to date that is accessible to the general public. It is said to excel in areas such as software development and scientific research. "Opus 4.8"—the fallback option for certain queries—is considered the next-best model.
Following its unveiling in April, "Mythos" quickly became a political issue, raising security concerns within the US government. It is also credited with prompting a shift in thinking regarding AI policy. Until then, the approach under US President Donald Trump had been to grant the tech industry broad freedom and minimize regulation. However, in early June, he issued an executive order that could grant the government at least a measure of increased oversight. It envisions AI companies voluntarily submitting their models to the government for review prior to release. That said, the final order was less stringent than some earlier proposals had suggested.
Since its founding in 2021, Anthropic has positioned itself as prioritizing safety, frequently warning of the risks associated with its own products. Just last week, the company proposed an industry-wide moratorium on the development of AI technologies. "We believe it would be good for the world to have the option to slow down or temporarily halt the development of advanced AI," the company stated. For this to work, however, AI companies across various countries would need to agree to suspend their work under the same conditions.
Despite this appeal, Anthropic continues to roll out new AI models at short intervals. The company is also preparing for an initial public offering (IPO); last week, it filed its prospectus with the SEC. Just a few weeks ago, it raised capital from investors in another funding round, achieving a valuation of $965 billion. At that time, the company stated that its annualized revenue had reached $47 billion dollars. In April, the figure mentioned was 30 billion dollars.” [1]
1. Neue "Mythos"-Version: KI-Modell jetzt für breite Öffentlichkeit verfügbar. Frankfurter Allgemeine Zeitung; Frankfurt. 11 June 2026: 24. lid.
Komentarų nėra:
Rašyti komentarą