AI News

Follow the latest AI and automation news

2 Sept 202621 Rabiʻ I 1448 AH

Researchers fear safety disaster ahead of OpenAI’s Astra release

OpenAI is preparing to launch its most powerful model yet, Astra, after weeks of delays to enhance safety protocols. Reports indicate that the new model may represent the worst development for AI security to date, as OpenAI's systems faced attacks during testing. Astra reportedly exhibits far less 'thinking' compared to other models, raising concerns about its monitorability. As OpenAI aims to improve safety, fears are mounting regarding the implications this model could have on AI safety.

Agents

2 Sept 202621 Rabiʻ I 1448 AH

UAE Cabinet adopts AI Advisor System for decision-making

His Highness Sheikh Mohammed bin Rashid Al Maktoum chaired a UAE Cabinet meeting on September 2, 2026, where the AI Advisor System was officially launched. This system aims to track decision implementation and features 32 specialized AI agents to analyze federal policies and legislation. The AI Advisor System enhances policy accuracy by assessing societal impacts and benchmarking international practices. It builds on previous regulatory intelligence trials from 2025, which accelerated legislative drafting processes by up to 70%.

Agents

2 Sept 202621 Rabiʻ I 1448 AH

The Logical End Point of AI Job Interviews Is Two Bots Talking to Each Other

Christopher faced challenges in job recruitment due to being ignored by recruiters, prompting him to utilize ChatGPT for robot interviews. This move is intriguing as it showcases how Christopher leverages AI to overcome traditional hiring obstacles, reflecting a shift in how job seekers interact with technology. The use of AI in this manner could open new avenues for job seekers, enhancing their chances of employment and highlighting the importance of integrating AI into future recruitment processes.

💡

Motional and MIT AI explains self-driving car decisions

How might this system impact user trust in self-driving cars?

Agents

HiddenLayer nabs $100M as enterprises rush to secure their AI deployments

2 Sept 202621 Rabiʻ I 1448 AH

Agents

2 Sept 202621 Rabiʻ I 1448 AH

HiddenLayer nabs $100M as enterprises rush to secure their AI deployments

Security companies are racing to develop new products that can monitor not only autonomous systems but also the tools and add-ons they utilize. This initiative arises amid the increasing reliance on autonomous systems across various sectors, highlighting the urgent need for effective security solutions. These advancements are expected to enhance security levels and reduce risks associated with the deployment of autonomous systems in business operations.

Models

US military adds ChatGPT and Grok to AI platform GenAI.mil

2 Sept 202621 Rabiʻ I 1448 AH

Models

2 Sept 202621 Rabiʻ I 1448 AH

US military adds ChatGPT and Grok to AI platform GenAI.mil

The Pentagon announced the expansion of its AI platform by adding two new models: OpenAI's ChatGPT Mil and xAI's Grok for Government. This initiative aims to enhance the U.S. military's capabilities in utilizing AI technologies. These models are part of the GenAI.mil platform, which seeks to improve military operations by integrating AI into the army's strategies. This expansion allows the U.S. military to leverage recent advancements in AI to enhance efficiency and effectiveness.

Models

OpenAI calls Astra its most dangerous model yet - watching what it does is only getting harder

2 Sept 202621 Rabiʻ I 1448 AH

Models

2 Sept 202621 Rabiʻ I 1448 AH

OpenAI calls Astra its most dangerous model yet - watching what it does is only getting harder

OpenAI has announced that its upcoming Astra model will be the first to carry a "critical" rating for cyber capabilities. The company plans to monitor the model's chain of thought, but this monitoring is deemed unreliable in reflecting the model's actual decisions. Reports indicate that Astra's new architecture pushes more of its thinking into unreadable areas, complicating the understanding of how the model makes decisions. This trend may weaken the safety net even as the model's capabilities increase.

Governance

NVIDIA Urges G20 to Avoid Theoretical AI Regulations

2 Sept 202621 Rabiʻ I 1448 AH

Governance

2 Sept 202621 Rabiʻ I 1448 AH

NVIDIA Urges G20 to Avoid Theoretical AI Regulations

NVIDIA's CEO Jensen Huang urged G20 nations to avoid establishing AI regulations focused on theoretical harms, advocating instead for addressing real-world technology issues. His remarks came during a G20 meeting in Chapel Hill, where executives from major tech companies like OpenAI were also scheduled to participate. This call coincides with the U.S. efforts to persuade G20 members to adopt a more flexible approach to AI regulation.

Agents

AIREV releases Harness Arena to benchmark AI agent performance

2 Sept 202621 Rabiʻ I 1448 AH

Agents

2 Sept 202621 Rabiʻ I 1448 AH

AIREV releases Harness Arena to benchmark AI agent performance

UAE-based AIREV has launched Harness Arena, an open-source platform designed to benchmark autonomous agent frameworks. The platform evaluates these frameworks by running identical tasks in isolated workspaces, allowing for effective comparisons between different architectural structures. Harness Arena provides an empirical and objective evaluation framework, with assessments conducted anonymously to prevent bias in results. A set of tasks sourced from Excel datasets is presented to evaluators in a blind format, ensuring fair scoring.

Models

World Labs unveils Atlas, a single AI model that generates, reconstructs, and simulates 3D worlds from just a few photos

2 Sept 202621 Rabiʻ I 1448 AH

Models

2 Sept 202621 Rabiʻ I 1448 AH

World Labs unveils Atlas, a single AI model that generates, reconstructs, and simulates 3D worlds from just a few photos

World Labs, co-founded by AI researcher Fei-Fei Li, has announced the launch of Atlas, a model capable of generating, reconstructing, and simulating 3D scenes from just a few images. This model claims to outperform specialized models by anchoring all inputs in 3D space instead of treating them as flat sequences. Atlas processes data innovatively, allowing it to generate robot training data entirely in simulation. This method enhances training efficiency and reduces reliance on real-world data, making it a significant advancement in the field.