AI News

Follow the latest AI and automation news

11 Sept 202630 Rabiʻ I 1448 AH

Anthropic Reveals AI Models Used in Biological Weapons Research

Anthropic reported five instances where scientists used its AI models, including Claude, for research related to developing biological weapons. The company halted these activities after confirming their use for unethical purposes. The report highlights the security measures taken to prevent its models from being used in biological weapons development, reflecting its commitment to ethical responsibility. These steps come amid growing concerns about the negative uses of technology.

Agents

11 Sept 202630 Rabiʻ I 1448 AH

Google Launches Gemini App for Windows to Compete with ChatGPT and Copilot

Google has officially launched the Gemini app for computers running Windows 10 and 11. This app aims to compete directly with AI applications like ChatGPT and Copilot, after a period of relying mainly on web experiences. The app is now globally available for Windows devices, supporting both x64 and ARM64 processors. This launch represents a strategic move for Google to strengthen its presence in the AI application market, allowing users direct interaction with the intelligent assistant.

Governance

11 Sept 202630 Rabiʻ I 1448 AH

New Mexico Court Penalizes Lawyer for Using AI in Murder Case

The New Mexico Supreme Court fined attorney Stephen Aarons for submitting an appeal that included false testimony from fabricated witnesses created by ChatGPT. The court deemed Aarons showed a lack of concern for his client, imposing a $5,000 fine and referring him to the attorney disciplinary board. Aarons used AI to summarize trial proceedings but failed to recognize the potential for inaccuracies. He expressed regret but hopes the disciplinary board will consider the circumstances surrounding his mistake.

💡

Palantir Foundry and cuOpt drive NVIDIA supply chain allocation

How might this technology impact supply chains in other companies?

Agents

Alteryx Launches New Capabilities to Enhance AI in Business

11 Sept 202630 Rabiʻ I 1448 AH

Agents

11 Sept 202630 Rabiʻ I 1448 AH

Alteryx Launches New Capabilities to Enhance AI in Business

Alteryx has announced the launch of new capabilities within the Alteryx One platform, connecting enterprise business logic directly with AI agents. These capabilities allow for the expansion of workflows and managed datasets to include external AI tools, enhancing operational effectiveness. The new features rely on a 'build once, manage once' approach, reducing the need for rework and increasing resource efficiency. This method enables users to seamlessly integrate AI tools without the need to redevelop solutions.

Governance

Veeam warns of EMEA ‘shadow agent’ crisis

11 Sept 202630 Rabiʻ I 1448 AH

Governance

11 Sept 202630 Rabiʻ I 1448 AH

Veeam warns of EMEA ‘shadow agent’ crisis

New research from Veeam reveals that 70% of organizations admit automated AI workflows interact with sensitive data without full oversight. Organizations face dual challenges in AI management, with 67% reporting employees create autonomous workflows that IT cannot fully track. This highlights the urgent need for ensuring data security and trustworthiness. As regulatory pressures increase, AI governance has become a C-suite liability, with 40% of leaders concerned about personal accountability. Additionally, 39% of organizations report heightened scrutiny from boards, increasing tension among executives.

Models

Rapidly scaling online storage to serve over 1 billion ChatGPT users

11 Sept 202630 Rabiʻ I 1448 AH

Models

11 Sept 202630 Rabiʻ I 1448 AH

Rapidly scaling online storage to serve over 1 billion ChatGPT users

OpenAI has announced the evolution of Habitat, which has transitioned from a Python library to a global storage platform. The platform now supports one billion ChatGPT users and handles 22 million requests per second. Habitat exemplifies how cloud storage performance can be enhanced, designed to meet the growing demands of users. This development reflects OpenAI's commitment to providing effective and scalable solutions in the AI landscape.

Models

The Mathematical AI Safety Institute wants to prove AI is safe the way cryptographers prove codes are unbreakable

11 Sept 202630 Rabiʻ I 1448 AH

Models

11 Sept 202630 Rabiʻ I 1448 AH

The Mathematical AI Safety Institute wants to prove AI is safe the way cryptographers prove codes are unbreakable

Canadian mathematician Jacob Tsimerman, a recent Fields Medal recipient, has announced the founding of the Mathematical AI Safety Institute (MAISI). The institute aims to demonstrate AI safety in a manner similar to how cryptographers prove that codes are unbreakable. This initiative represents a significant advancement in the AI field, focusing on developing new methodologies to ensure the safety of intelligent systems. By employing advanced mathematical techniques, Tsimerman and his team hope to provide a reliable framework for assessing the risks associated with AI technologies.

Agents

Why So Many AI Researchers Think the Machines Could Kill Everyone

11 Sept 202630 Rabiʻ I 1448 AH

Agents

11 Sept 202630 Rabiʻ I 1448 AH

Why So Many AI Researchers Think the Machines Could Kill Everyone

Major AI labs are experiencing rapid technological advancements that are causing concern among researchers. These developments include recursive self-improvement and the use of agentic swarms, complicating the landscape further. These new dynamics present significant challenges for researchers, as the rapid progress may lead to unexpected outcomes affecting safety and reliability.

Agents

OpenAI's new Agents API gives developers the infrastructure behind Codex and ChatGPT

11 Sept 202630 Rabiʻ I 1448 AH

Agents

11 Sept 202630 Rabiʻ I 1448 AH

OpenAI's new Agents API gives developers the infrastructure behind Codex and ChatGPT

OpenAI has announced the release of its Agents API as a public beta, allowing developers to create cloud agents that operate autonomously. These agents can execute code and delegate tasks to sub-agents, enhancing developers' ability to build more complex applications. The new API operates without additional fees beyond token usage, making it widely accessible to developers. Additionally, companies like Cloudflare, Vercel, and Oracle provide extra sandbox environments, facilitating developers' experimentation with this technology.