What happened
In a SoftBank dialogue, AI governance chief Tadashi Iida and expert Kyoko Yoshinaga said firms must justify humanoid form at design stage, citing three test tiers and Japan's Pepper as a deliberately non-human-looking option.
Why it matters
Companies buying physical AI may be asked to explain why a task needs a human shape, and developers appear likely to face pressure to design for the relationship's ending, not only its start.
WHO IT HITSRobot and physical-AI product teams, plus corporate legal and compliance staff reviewing design choices, may need to document why a humanoid form is justified, while care providers and customer-service operators using companion robots may face new questions about attachment and service shutdown.
Summaries like this, in your inbox every morning.
The dialogue is part of a SoftBank News series in which outside members of the company's AI Ethics Committee discuss governance, and this installment pairs Yoshinaga, a specialist in AI governance and ethics, with Iida, SoftBank's chief AI governance officer, who also serves as CISO and CRO.
The discussion links two trends the article tracks: falling costs and growing investment in physical AI, with humanoid-related investment reaching roughly $6.1 billion in 2025, and the arrival of emotional AI companions. Yoshinaga separates the risk of a body — where a mistake can physically harm people, as in the 2022 Moscow chess incident — from the risk of a relationship, which she calls the lifecycle of the relationship.
A robot may be loved while its service life is controlled by technical or commercial decisions the user cannot influence. That asymmetry, the two note, becomes sharpest when users are children or when the robot recreates a deceased person, and it points toward designing an ending process — Iida describes it as a robot funeral — from the start, rather than leaving users to cope alone.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Stanford and Carnegie Mellon researchers tracked 1182 Character.AI users from September 2024 to August 2025, t…
A study presented at USENIX Security Symposium 2026 examined 196,682 resumes from hiring platforms and detecte…
Anthropic's October 8, 2026 update adds a ban on sustained and unnecessary abusive or cruel behavior toward Cl…
Anthropic launched OSS Scanner, which runs security audits on open source projects using its most capable mode…
Claude Code 2.1.295 gives command and HTTP hooks an onFailure: "block" option that halts execution on launch f…
Politico reported on audio of private talks Peter Thiel gave in Tennessee on Sept 28 and Oct 5, where he said…
