Meta Assistant Muse Exposed to Zero-Day Vulnerability Allowing Account Theft · 0922-673

2026-09-22 08:00

Author: RadarAI Editorial Editor: RadarAI Editorial Last updated: 2026-09-22 Review status: Editorial review pending Brief 速报官方 AI动态开源

Editorial standards and source policy: Editorial standards, Team. Content links to primary sources; see Methodology.

## 🔍 Core Insights **Meta**'s AI assistant **Muse** became the focus this week: on one hand, it partnered with **Shopify** to integrate **Shop Pay** and enable agentic shopping [1][24]; on the other hand, it was exposed to a severe **zero-day vulnerability** in which a local app can steal account tokens [18]. Meanwhile, **Xiaomi** open-sourced the **MiMo-V2.6** series models [8][17], which multiple KOLs evaluated as surpassing **Grok 4.7** at lower cost [23], while **OpenAI** has basically achieved automation of the training process for new experimental models internally [13]. ## 🚀 Key Updates - **Meta Muse integrates with Shopify Shop Pay** [1][24]: The Muse agent can complete agentic checkout in Shopify stores; Amazon had previously blocked the agent. - **Meta Muse exposed to severe zero-day vulnerability** [18]: A local app can steal authentication tokens by changing the voice transcription endpoint, gaining full control of the account. - **Xiaomi open-sources MiMo-V2.6 series** [8][17]: Includes Pro and Flash versions; Pro scored 46 on the AA Intelligence Index v4.3.2, with API pricing unchanged from the previous generation. - **Grok 4.7 released with outstanding cost-effectiveness** [11][12]: On AA-Briefcase, it ranks second only to Anthropic models, with a per-task cost about 50% of Opus 5. - **OpenAI's internal training process is basically automated** [13]: The new model solved over 100 open math problems, surprising researchers as well. - **AI glasses shipments up 263% year over year** [2]: Counterpoint data shows screenless products dominate, growing 258% year over year. - **UK Citizens Advice calls for an end to customer service reliance on AI** [5]: It demands that essential service companies in energy, banking, and other sectors guarantee users' right to communicate with a human. - **Linear reviews CI bottlenecks after AI coding acceleration** [4]: The test suite grew nearly fourfold during the year, while PR CI wait time dropped from just over 6 minutes to just over 5 minutes. ## 🔗 Sources [1] Meta partners with Shopify; Muse will enable Shop Pay checkout in Shopify stores — https://aihot.news/items/cmubw42we058wro9968uzn2te [2] Counterpoint data shows global AI glasses shipments in H1 2026 grew 263% year over year, with screenless products dominating — https://aihot.news/items/cmubvi9a804ofro9992w0m6f1 [4] Linear engineering retrospective: how to rebuild CI bottlenecks after AI coding acceleration — https://aihot.news/items/cmubvcztx04ifro99anhzxxt1 [5] UK Citizens Advice calls on essential service companies to stop relying on AI chatbots for customer service — https://aihot.news/items/cmubvi9a804ogro993qhqo004 [8] Xiaomi releases and open-sources the MiMo-V2.6 series, with Pro and Flash versions, API pricing unchanged from the previous generation — https://aihot.news/items/cmubvi9a804ohro99sfsvkd9m [11] Grok 4.7 ranks second only to Anthropic models on AA-Briefcase, closely trailing Opus 5, with per-task cost about 50% of it — https://aihot.news/items/cmubvcpe404icro99u1tpb7sv [12] SpaceXAI releases Grok 4.7, focused on coding and knowledge processing, starting at $2 per million tokens input — https://aihot.news/items

Meta's AI assistant Muse became the focus this week: on one hand, it partnered with Shopify to integrate Shop Pay and enable agentic shopping [1][24]; on the other hand, it was exposed to a severe zero-day vulnerability in which a local app can steal account tokens [18]. Meanwhile, Xiaomi open-sourced the MiMo-V2.6 series models [8][17], which multiple KOLs evaluated as surpassing Grok 4.7 at lower cost [23], while OpenAI has basically achieved automation of the training process for new experimental models internally [13].

🚀 Key Updates

Meta Muse integrates with Shopify Shop Pay [1][24]: The Muse agent can complete agentic checkout in Shopify stores; Amazon had previously blocked the agent.
Meta Muse exposed to severe zero-day vulnerability [18]: A local app can steal authentication tokens by changing the voice transcription endpoint, gaining full control of the account.
Xiaomi open-sources MiMo-V2.6 series [8][17]: Includes Pro and Flash versions; Pro scored 46 on the AA Intelligence Index v4.3.2, with API pricing unchanged from the previous generation.
Grok 4.7 released with outstanding cost-effectiveness [11][12]: On AA-Briefcase, it ranks second only to Anthropic models, with a per-task cost about 50% of Opus 5.
OpenAI's internal training process is basically automated [13]: The new model solved over 100 open math problems, surprising researchers as well.
AI glasses shipments up 263% year over year [2]: Counterpoint data shows screenless products dominate, growing 258% year over year.
UK Citizens Advice calls for an end to customer service reliance on AI [5]: It demands that essential service companies in energy, banking, and other sectors guarantee users' right to communicate with a human.
Linear reviews CI bottlenecks after AI coding acceleration [4]: The test suite grew nearly fourfold during the year, while PR CI wait time dropped from just over 6 minutes to just over 5 minutes.

🔗 Sources

[1] Meta partners with Shopify; Muse will enable Shop Pay checkout in Shopify stores — https://aihot.news/items/cmubw42we058wro9968uzn2te [2] Counterpoint data shows global AI glasses shipments in H1 2026 grew 263% year over year, with screenless products dominating — https://aihot.news/items/cmubvi9a804ofro9992w0m6f1 [4] Linear engineering retrospective: how to rebuild CI bottlenecks after AI coding acceleration — https://aihot.news/items/cmubvcztx04ifro99anhzxxt1 [5] UK Citizens Advice calls on essential service companies to stop relying on AI chatbots for customer service — https://aihot.news/items/cmubvi9a804ogro993qhqo004 [8] Xiaomi releases and open-sources the MiMo-V2.6 series, with Pro and Flash versions, API pricing unchanged from the previous generation — https://aihot.news/items/cmubvi9a804ohro99sfsvkd9m [11] Grok 4.7 ranks second only to Anthropic models on AA-Briefcase, closely trailing Opus 5, with per-task cost about 50% of it — https://aihot.news/items/cmubvcpe404icro99u1tpb7sv [12] SpaceXAI releases Grok 4.7, focused on coding and knowledge processing, starting at $2 per million tokens input — https://aihot.news/items

← Back to Updates