Daily AI News | Alibaba Cloud adds 12 agent tools to Token Plan Personal without changing prices or credits
# Daily AI News | Alibaba Cloud adds 12 agent tools to Token Plan Personal without changing prices or credits
## Lead Story
### Alibaba Cloud adds 12 agent tools to Token Plan Personal without changing prices or credits
Alibaba Cloud added 12 Agent Harness tool categories to the Standard and Pro tiers of Token Plan Personal while keeping plan prices and credit allocations unchanged. The tools cover web search and parsing, images, speech, code, and related agent capabilities; the official plan page still lists seven-day prices of RMB 39, RMB 139, and RMB 499 for Lite, Standard, and Pro.
**AI take:** The update bundles integrations that developers might otherwise procure separately. Its real value depends on tool-call credit consumption and whether a team already pays for overlapping services.
## More News
### Ant Group open-sources HOP 3.0 for bounded and verifiable agent execution
Ant Group's privacy-computing team announced the open-source release of HOP 3.0. The release centers on a trusted-native agent language, controllable execution boundaries, and verifiable processes intended to let agents explore while keeping critical actions constrained, auditable, and traceable.
**AI take:** Open sourcing lowers the cost of evaluating this approach. Production adoption will still depend on permission isolation, audit completeness, and failure recovery in real tasks.
### Ant Group launches APASS trust infrastructure for agent commerce
Ant Group launched APASS, trust infrastructure for commercial activity performed by agents. Based on a Know Your Agent approach, it builds identity and behavioral trust chains with registration, continuous verification, intent security, and tamper-resistant records for service calls, business processes, and payments.
**AI take:** Identity and authorization become foundational once agents transact for people. Adoption across platforms, false-block rates, and accountability will determine whether APASS works in practice.
### JoyIn announces its Aether humanoid-control model and plans to open-source selected components
CNBC reported that JoyIn publicly announced Aether, a humanoid-control model built around perception rather than primarily text. The company claims a 90% first-attempt task success rate and a two-thirds reduction in training time, and plans to open-source selected components such as touch; those performance figures have not been independently verified.
**AI take:** Partial open sourcing can help developers evaluate the approach, but vendor metrics do not replace field testing. Stability and reproducibility across different robot hardware are the key tests.
### ChatGPT upgrades Sites for creating, sharing, and hosting websites in one flow
OpenAI product lead Tibo announced an upgrade to Sites in ChatGPT, allowing users to create, preview, share, and host websites from ChatGPT. OpenAI's help documentation notes that public publishing, external sharing, and organizational access remain subject to workspace settings and supported availability.
**AI take:** The feature shortens the path from an idea to a reachable page. Organizations should still confirm publishing permissions, data boundaries, and who maintains generated code.
### Git AI team members join OpenAI to continue work on coding-agent attribution
OpenAI product lead Tibo confirmed that Aidan and Sasha from the Git AI team have joined OpenAI. Their open-source work helps organizations identify and understand contributions made by coding agents; no specific product-integration schedule was announced in the window.
**AI take:** Attribution affects code review and accountability. The impact depends on whether this reaches Codex or enterprise tools.
## Rumors
### Nvidia reportedly discusses investing up to $10 billion in a potential Anthropic IPO
Reuters, citing two people familiar with the discussions, reported that Nvidia is in talks to become an anchor investor in a potential Anthropic IPO with an investment of up to $10 billion. Anthropic is reportedly considering raising as much as $100 billion at a valuation near $2 trillion, but the discussions may change and neither company confirmed an agreement in the window.
**AI take:** If it proceeds, the capital relationship between a frontier-model company and a critical compute supplier would deepen. Formal filing terms, the actual investment, and supply arrangements are the next evidence to watch.
