AI News

Follow the latest AI and automation news

23 May 20267 Dhuʻl-Hijjah 1447 AH

Deep Seek Announces Permanent 75% Price Cut for V4-Pro Model

Deep Seek, a Chinese AI startup, announced a permanent 75% price reduction for its flagship V4-Pro model, lowering prices to a quarter of their original levels. The company did not clarify if this reduction is linked to increased supplies of Huawei's Ascend 950 chips, which are crucial for enhancing the model's performance. The API costs for the V4-Pro now range from 0.025 to 6 yuan per million tokens, down from 0.1 to 24 yuan. The token serves as the basic unit for measuring the data processed by the AI model, making this price cut significant for competitive pricing in the market.

Agents

23 May 20267 Dhuʻl-Hijjah 1447 AH

Alibaba's latest AI model ran autonomously for 35 hours to optimize code for its own custom chip

Alibaba's Qwen team has launched the Qwen3.7-Max model, designed for long-term autonomous agent tasks. This model outperforms Chinese competitors like DeepSeek V4 Pro and Kimi K2.6, matching Claude Opus 4.6 in benchmarks. The team demonstrated the model's ability to control a quadruped robot, showcasing its potential for practical applications. The Qwen3.7-Max model optimizes code for a custom chip, having run autonomously for 35 hours. This achievement reflects significant advancements in artificial intelligence and the model's capability to handle complex tasks.

Governance

23 May 20267 Dhuʻl-Hijjah 1447 AH

Anthropic warns Claude Mythos Preview finds bugs faster than developers can patch them

Anthropic revealed that its Claude Mythos model, collaborating with around 50 partners under the Glasswing project, has identified over 10,000 critical vulnerabilities in core software. These vulnerabilities are increasing at a rate that exceeds developers' ability to patch them, creating a high-risk transitional period. Anthropic warns that no company, including itself, has implemented sufficient security measures to prevent the misuse of these models. This situation raises concerns in the AI sector, necessitating companies to strengthen their security strategies.

💡

[AINews] All Model Labs are now Agent Labs

What do you think about OpenAI's shift towards agent development?

Governance

AI is being used to resurrect the voices of dead pilots

22 May 20266 Dhuʻl-Hijjah 1447 AH

Governance

22 May 20266 Dhuʻl-Hijjah 1447 AH

AI is being used to resurrect the voices of dead pilots

Researchers have utilized AI on cockpit recordings to reconstruct the audio content. This development prompted the U.S. National Transportation Safety Board (NTSB) to temporarily restrict access to its file system. This move raises controversial questions regarding privacy and security in using AI for analyzing sensitive data. It also highlights the challenges regulatory bodies face in addressing technological innovations. The incident underscores the need for clear policies on AI use in sensitive areas, particularly in aviation, urging decision-makers to balance innovation with the protection of sensitive information.

Agents

Did Google’s AI agents really build an operating system for $916?

22 May 20266 Dhuʻl-Hijjah 1447 AH

Agents

22 May 20266 Dhuʻl-Hijjah 1447 AH

Did Google’s AI agents really build an operating system for $916?

Google recently launched its new model, Gemini 3.5 Flash, at the latest developer conference, alongside the Antigravity 2.0 application. The company claimed that a team of agents built a complete operating system with just one request, costing $900 in API fees. However, the assertion that it was a 'single' request is misleading, as it was revealed to consist of thousands of lines. The process involved a specialized framework with designated roles and delegation to sub-agents, raising questions about the effectiveness of this structure for other complex programming tasks.

Agents

OpenAI launches a ChatGPT Powerpoint plugin and warns it might accidentally delete your content

22 May 20266 Dhuʻl-Hijjah 1447 AH

Agents

22 May 20266 Dhuʻl-Hijjah 1447 AH

OpenAI launches a ChatGPT Powerpoint plugin and warns it might accidentally delete your content

OpenAI has launched a new ChatGPT plugin for PowerPoint that allows users to create presentations from notes, documents, or images. This add-in is available globally across all subscription tiers, enhancing the user experience in presentation creation. OpenAI advises users to save important presentations before using the plugin, as it may unintentionally delete content during certain operations. This initiative reflects OpenAI's commitment to integrating its technologies into various productivity tools.

Business

Deepseek reportedly prioritizes AGI research over quick profits despite billions in funding

22 May 20266 Dhuʻl-Hijjah 1447 AH

Business

22 May 20266 Dhuʻl-Hijjah 1447 AH

Deepseek reportedly prioritizes AGI research over quick profits despite billions in funding

Chinese company Deepseek is preparing to raise approximately $10 billion, increasing its valuation to around $45 billion. Founder Liang Wenfeng has informed investors that he prioritizes general artificial intelligence research over quick profits. This bold move comes amid fierce competition in the AI market, where many companies are focused on rapid returns. However, Liang believes that investing in general AI research will yield long-term results that strengthen the company's market position.

Agents

OpenAI Appshots turn any Mac window into context for Codex

22 May 20266 Dhuʻl-Hijjah 1447 AH

Agents

22 May 20266 Dhuʻl-Hijjah 1447 AH

OpenAI Appshots turn any Mac window into context for Codex

OpenAI has launched its new Appshots feature for Mac users, allowing them to send the contents of any app window to the Codex coding assistant with a single button press. This feature provides Codex with the necessary context to perform tasks more effectively. Appshots simplifies user interaction with Codex, enabling direct sharing of information from any application, thus enhancing programming efficiency. This advancement marks a significant step in the integration of AI into software development.

Models

We tried Google’s AI glasses and they’re almost there

22 May 20266 Dhuʻl-Hijjah 1447 AH

Models

22 May 20266 Dhuʻl-Hijjah 1447 AH

We tried Google’s AI glasses and they’re almost there

Google has unveiled a prototype of XR glasses powered by Android, utilizing Gemini technology for instant translation. These glasses provide real-time information within the user's field of vision, facilitating navigation and interaction with digital content. This innovation marks a significant step in integrating artificial intelligence with wearable devices, enhancing the user experience in daily life. The glasses merge digital information with the real world, allowing users to access translation and navigation instantly.