In recent months, major artificial intelligence (AI) companies have made high-profile announcements, often claiming groundbreaking achievements. Anthropic, a leading AI firm, stated that its model, Claude Mythos, is more effective at identifying software vulnerabilities than many human security experts. This was followed by a hacking incident involving OpenAI and Hugging Face, two other prominent AI companies. Anthropic and Meta later revealed similar incidents involving their AI models. Anthropic also claimed a mathematical breakthrough, which was quickly followed by a similar claim from OpenAI. Recently, an Anthropic engineer named Jacob Coxon left the company, warning that both Anthropic and OpenAI are "racing straight towards self-improving superintelligence and gambling with our lives." These developments have been widely covered by the media, often using language that suggests AI systems are approaching "artificial general intelligence," a hypothetical form of AI capable of performing any intellectual task a human can. However, cybersecurity experts argue that the "hacking" incidents primarily highlight OpenAI’s negligence and failure to implement basic security practices, rather than demonstrating AI models acting independently or maliciously. Regarding the mathematical breakthroughs, mathematicians were initially surprised by OpenAI’s claim that its chatbot, Astra, solved long-standing problems. However, many later found the results to be less novel than claimed. Some mathematicians accused OpenAI of research misconduct and plagiarism, noting that Astra did not achieve a "profound intellectual leap." Soon after, Tristan Buckmaster, a math professor at New York University, suggested that OpenAI had stolen other people’s work. Claims of dangerous "superintelligence" are often criticized as narratives shaped more by ideology than by scientific or engineering evidence. The focus on fields like computer programming and mathematics, where problems often have clear right or wrong answers, may be a strategic move by AI companies to highlight their models’ capabilities. These areas are also seen as representing the pinnacle of human intellectual achievement, which can make AI advancements seem more impressive. Mathematicians have warned against corporations overstating the capabilities of their AI products and have urged policymakers to consult with experts rather than relying on press releases. They argue that the tech industry’s emphasis on urgency and speed is a way to misdirect attention from real issues. By framing AI as "superintelligence" or "rogue models," companies may be shifting public focus away from their own accountability, such as concerns about OpenAI potentially creating malware or plagiarizing academic work. Instead, attention is redirected toward fears about fictional superintelligent machines. The AI industry has also suggested that activism against data centers—large facilities that house servers and consume significant energy—is a distraction from regulating "superhuman" AI. However, concerns about the environmental impact, health effects, and resource use of data centers remain valid and are not easily dismissed. Experts stress the importance of skepticism and independent expert input in policymaking, urging policymakers and the public to recognize corporate hype for what it is.