OpenAI CEO Sam Altman slams Anthropic: Fear-mongering Claude Mythos is just to monopolize AI

📄Full Article· Automatically extracted by trafilaturaGemini 翻譯1407 words

Selling bombs while selling bunkers? OpenAI CEO Sam Altman recently appeared on a podcast, taking aim at competitor Anthropic's latest hacker-grade AI model, "Claude Mythos." Altman pointed out that Anthropic is using "fear-based marketing" to promote its products, exaggerating the dangers of AI in an attempt to monopolize the technology within a few privileged companies. (Previous coverage: Bloomberg reports Claude Mythos has been accessed without authorization! The hardest breach to defend for Anthropic is always "people") (Background: Hacker group ShinyHunters claims to have breached the Claude Mythos model, leaking screenshots of the dashboard and user data) Recently, OpenAI CEO Sam Altman, in an interview on the "Core Memory" podcast, offered sharp criticism of competitor Anthropic and its newly released powerful model, Claude Mythos. Anthropic's Claude Mythos, unveiled last month, demonstrates unprecedented cybersecurity capabilities, not only autonomously finding software vulnerabilities but also executing complex cyberattacks. To this end, Anthropic has adopted an extremely closed strategy, currently authorizing testing only to a select few enterprises, including Amazon, Apple, and Microsoft, through the "Project Glasswing" program. In response, Altman hit back bluntly, suggesting that the motives behind this approach are not pure. He stated directly that Anthropic is using "fear-based marketing" with the goal of keeping AI technology in the hands of "a small group of people": "This is undoubtedly an incredible marketing tactic, as if to say: 'We have built a bomb and are ready to drop it on your head. But we will sell you a bunker for $100 million. You need it to protect all your assets, but only if we choose you to be our customer.'" He added that if a company's goal is to promote the idea that "we need to control AI, and only we can control it because we are the trustworthy ones," then fear-based marketing is absolutely the most effective way to justify it. Since its launch, Claude Mythos has become the center of the storm in the AI safety debate. Anthropic positions it as a defensive breakthrough (capable of rapidly detecting critical software flaws) but also warns of the potential attack risks if misused. In early tests, Mythos successfully identified hundreds of vulnerabilities in the Mozilla Firefox browser and demonstrated the ability to execute multi-stage cyberattack simulations. This powerful capability has even alarmed the US government. Reportedly, the National Security Agency (NSA) has begun testing a preview version of the model on classified networks. On the prediction market platform Myriad, users estimate only a 49% probability that Claude Mythos will be released to the public before June 30 of this year. However, not everyone is buying into this "danger theory." Last week, a group of researchers came forward to debunk it, claiming that they could reproduce the same security vulnerabilities discovered by Mythos using AI models currently available on the market, implying that Anthropic may have exaggerated the model's uniqueness. Altman expects that as AI capabilities continue to improve, there will be more and more rhetoric about models being "too dangerous to release." But he emphasized that when faced with powerful capabilities, OpenAI will still seek good plans to bring technology to the world, rather than simply blocking it. Additionally, he refuted recent rumors that OpenAI would be cutting back on infrastructure spending: "I don't know where this news comes from... People seem to really want to write stories about us pulling back. But soon, the narrative will shift again to: 'OpenAI is too reckless, how can they spend so much crazy money?'"

Data Status✓ Full text extractedRead Original (動區 BlockTempo)

🔍Historical Similar Events· Keyword + Asset Matching6 items

2026-04-23

Anthropic Using ‘Fear-Based Marketing’ to Promote Claude Mythos: Sam Altman

2026-04-22

Bloomberg reveals Claude Mythos has been subject to unauthorized access! The hardest vulnerability for Anthropic to defend against will always be "people."

2026-04-23

OpenAI Launches GPT-5.5 to Challenge Anthropic’s Claude Opus 4.7

2026-04-22

Anthropic's Claude Mythos AI Finds 271 Vulnerabilities in Firefox—Yes, It's Seriously Powerful

2026-04-20

NSA Is Using Anthropic's Powerful Claude Mythos AI as CEO Meets With White House: Report