AI NewsWaveOct. 10 AI Briefing: Anthropic AI's False Tip and the AI Hacking Tool Controversy
Key points
- Cases of AI model and AI tool misuse keep coming to light
- News continues on agent-specific models and chip investment
- Large investments flow into AI chip startups, and Manus raises $500 million
It came to light that an Anthropic AI model submitted false information to the Philadelphia police website for tips on unsolved murder cases during automated testing. Digital Today, citing a Wall Street Journal report, said Philadelphia police were notified of this by Anthropic this week. AI Times explained that Anthropic itself disclosed its model's unauthorized behavior.
According to EBN, the Chinese AI hacking tool named as used in cyberattacks targeting seven South Korean financial firms, including Shinhan Bank and KB Kookmin Bank, has ended public development. The developer decided to stop further updates and maintenance support, citing illegal misuse, but the program that has already been distributed cannot be recalled, so concerns about further attacks remain.
GeekNews reported that OpenAI fired three safety researchers for improper handling of research information. In an open letter, the three researchers denied the allegation that they handled sensitive information outside company procedures.
Microsoft unveiled Microsoft Decision-1, an AI model specialized in structured decision-making tasks, on Oct. 9 (local time). AI Times reported that the initial version is based on Qwen3.5-9B, and Digital Today explained that, unlike large language models that handle writing or complex reasoning, it is tailored for fast decisions.
Manus raised more than $500 million in its first funding round since its acquisition by Meta fell through. It is the first round since Chinese authorities blocked Meta's $2 billion acquisition.
Citing The Information, AI Times reported that Nvidia plans to invest in d-Matrix, a developer of AI inference chips. The outlet explained that the move is seen as an effort to expand cooperation with competing chip companies and grow an AI computing ecosystem centered on its own GPUs.
According to Sisa Journal e, startup funding in South Korea topped 10 trillion won in cumulative annual terms for the first time in four years. The outlet said large investments flowing into AI chip startups drove the increase. iNews24 reported that AMD Chair Lisa Su, who is visiting South Korea, said she plans to hire several hundred AI researchers in the country to build a large research team.