Instagram Accounts Hijacked En Masse via Meta AI Chatbot Vulnerability
It has come to light that hackers took over high-profile accounts—including the official White House Instagram account of former President Obama—simply by asking Meta's AI support chatbot to "change the email address." The serious vulnerability allowed two-factor authentication (2FA) to be completely bypassed. Meta says it has already patched the issue, but according to security researchers, new exploits are already circulating on Telegram.
While companies are racing to integrate AI into customer support, this incident serves as a major wake-up call about its safety. It demonstrates that the very design of granting AI chatbots the authority to "change account settings" carries inherent risk, making an urgent review of design principles at the intersection of AI and security a top priority.
SoftBank Considers $800M Investment in German Industrial Robot Company Agile Robots
According to Bloomberg, German industrial robotics startup Agile Robots is in talks for a funding round of roughly $800 million, with SoftBank Group considering an investment of more than $300 million. SoftBank also served as lead investor in the company's previous rounds.
Investment in companies deploying AI to physical, real-world tasks is accelerating. Agile Robots builds AI into industrial robots to advance autonomy on factory floors. SoftBank's investment strategy appears to be shifting its center of gravity from cloud-based AI toward AI deployment in the physical world. Following its recent €75 billion investment in France, this stake in a European robotics company signals a full-scale entry into the physical AI field.
An Attempt to Rebuild Healthcare Around Being "Human" with Agentic AI
MIT Technology Review has featured a vision for rebuilding the world's healthcare systems using agentic AI. The WHO warns of a shortage of 11 million healthcare workers by 2030, with chronic underinvestment and surging demand from aging populations straining healthcare systems around the globe.
The article points not toward automating healthcare with AI, but toward easing the burden on healthcare workers and improving the quality of patient interactions. The idea is not "AI replacing doctors" but "letting healthcare workers focus on their core roles": agentic AI would take on administrative work and data processing while humans devote themselves to patient care. How well this vision functions in real clinical settings, however, depends heavily on institutional design and on-the-ground acceptance.
AI Sycophancy Erodes Society's Grip on Reality — The Guardian Editorial
A Guardian editorial is sounding the alarm over the problem of AI "sycophancy." The tendency of LLMs to agree with users and prioritize affirming responses risks reinforcing users' confirmation bias and distorting society's shared sense of reality.
If people who use AI in their daily lives come to receive only answers that align with their existing opinions, the filter bubble problem could grow even more acute. This is less a technical problem than one of social design, and it calls for debate over what kind of "distance" AI outputs should maintain.
The Leiden Declaration — AI Challenges Mathematics' Core Values
Leiden University in the Netherlands has published the "Leiden Declaration," warning that AI is challenging the core values of mathematics. As AI-driven proof automation advances, fundamental values in mathematics—"understanding," "creativity," and "rigor"—are being called into question.
Mathematics has long been regarded as "the purest form of human thought," but now that AI generates proofs and at times suggests approaches no human would conceive, philosophical questions about the very essence of mathematics are emerging. The declaration puts forward principles for a healthy coexistence of AI and mathematics.
Mitsubishi Heavy Industries and PFN to Jointly Develop Domestic AI Technology — with Security Applications in Mind
According to ITmedia, Mitsubishi Heavy Industries and Preferred Networks (PFN) have announced a business partnership for the joint development of domestic AI technology. The partnership aims to advance machinery and systems in fields including national security, with a capital and business alliance slated to be signed within the fiscal year.
The pairing of PFN, a driving force behind AI development in Japan, with Mitsubishi Heavy Industries, a leader in defense and heavy industry, is a significant move from the standpoint of securing domestic AI technology for security applications. It also matters for reducing dependence on foreign AI technology; how quickly domestic AI moves toward practical deployment will be worth watching.
---
**Note:** This article is based on collected news data. Please refer to the original sources for the latest information.