August 7, 2025

news  
meta, openai and google Updates

Several key developments are shaping the AI landscape. Australian arts, media, and voice actor groups are opposing the use of their content and voices to train AI models without consent or compensation, advocating for an "opt-in" system. Meanwhile, Y Combinator founder Paul Graham advises focusing on unique skills to avoid job displacement by AI, echoing Meta CEO Mark Zuckerberg's expectation that AI will soon write code at a mid-level engineer's capacity. On the hardware front, GPU architect Raja Koduri launched Oxmiq Labs to develop AI hardware and software, including OXCapsule, designed to enable AI and graphics to function across diverse hardware. This comes as tech giants are significantly increasing spending on cloud infrastructure to support AI, with Morgan Stanley projecting $359 billion in spending by major U.S. cloud providers in 2025. To meet the growing energy demands of AI, states are re-evaluating energy policies. In AI application news, Equinox is testing AI-driven workout and nutrition advice within its app, while SentinelOne acquired Prompt Security to enhance AI tool visibility for managed security service providers. However, IBM's research highlights the rising risk of data breaches due to inadequate AI oversight, with the average cost of a breach in the U.S. reaching $10.22 million.

Key Takeaways

  • Australian creatives are fighting for copyright protection against tech companies using their content to train AI models without permission.
  • The Australian Association of Voice Actors (AAVA) advocates for an "opt-in" system where individuals must consent to the use of their voices in AI training and be fairly compensated.
  • Y Combinator's Paul Graham advises focusing on specialized skills to protect jobs from AI, noting Meta CEO Mark Zuckerberg's prediction that AI will soon write code like a mid-level engineer.
  • Raja Koduri, a GPU architect, launched Oxmiq Labs to develop AI hardware and software, including OXCapsule, which allows AI and graphics to work on different hardware.
  • Morgan Stanley projects major U.S. cloud providers will spend $359 billion in 2025 on cloud infrastructure due to increased AI spending.
  • States are re-evaluating energy policies to meet the growing energy demands of AI.
  • Equinox is testing an AI feature in its app to provide personalized workout and nutrition advice.
  • IBM's study reveals that 16% of data breaches involve AI tools, with the average cost of a data breach in the U.S. reaching $10.22 million.
  • SentinelOne acquired Prompt Security to provide MSSPs with real-time visibility and policy enforcement for AI tools, helping to prevent data leaks and misuse.
  • Prompt Security detects Shadow AI, tools used without IT approval, and works with major AI models like OpenAI and Google.

Australian creatives fight tech firms using their content to train AI

Australian arts and media groups are pushing back against big tech companies using their content to train AI models without paying for it. A Productivity Commission report suggested changes to copyright law that could allow this, but the groups say it would be "rampant theft" of intellectual property. The Australian Council of Trade Unions (ACTU) also criticized the idea, saying it would hurt Australian workers. Attorney General Michelle Rowland says any changes must consider the impact on artists and news media.

Voice actors oppose AI training using their voices without permission

The Australian Association of Voice Actors (AAVA) is against companies using voices and images to train AI without consent. They worry about the lack of protection for people's identities and the potential for misuse of biometric data. AAVA wants an "opt-in" system where people must give permission first and be paid fairly. They believe current systems for licensing and paying creators should be used when AI competes with original work. AAVA President Simon Kennedy says everyone should have the right to control their identity in AI training.

Australia faces copyright fight over AI training data

A debate is happening in Australia about how to regulate AI training and copyright. A report suggests a possible "text and data mining exception" that would allow AI companies to use copyrighted material without permission. Critics like Jenny Morris from APRA say this would legitimize theft. Leah Flanagan from the National Aboriginal and Torres Strait Islander Music Office (Natsimo) is concerned about harm to Indigenous cultural property.

Y Combinator founder shares tips to protect your job from AI

Paul Graham, the founder of Y Combinator, says AI is taking over boring jobs. He says low-level programming jobs are already disappearing. Graham suggests focusing on your passions and doing something so well that AI can't compete. He also notes that Meta CEO Mark Zuckerberg expects AI to write code like a mid-level engineer soon. Other leaders like Mark Cuban and Jensen Huang believe AI will create new jobs too.

Y Combinator founder shares tips to protect your job from AI

Paul Graham, the founder of Y Combinator, says AI is taking over boring jobs. He says low-level programming jobs are already disappearing. Graham suggests focusing on your passions and doing something so well that AI can't compete. He also notes that Meta CEO Mark Zuckerberg expects AI to write code like a mid-level engineer soon. Other leaders like Mark Cuban and Jensen Huang believe AI will create new jobs too.

GPU legend Raja Koduri launches Oxmiq Labs for AI hardware and software

Raja Koduri, a well-known GPU architect, has started a new company called Oxmiq Labs. Oxmiq will create GPU hardware and software for AI and graphics. Their key product is OXCapsule, software that allows AI and graphics to work on different hardware. OXPython, a part of OXCapsule, lets Python-based CUDA applications run on non-Nvidia hardware. Tenstorrent's CEO Jim Keller is partnering with Oxmiq on their OXPython software stack.

AI spending drives cloud infrastructure surge among tech giants

Tech companies are spending more on AI, causing a big increase in cloud infrastructure. Morgan Stanley predicts that major U.S. cloud providers will spend $359 billion in 2025, a 57% jump. This growth is expected to continue into 2026. There's also a lot of demand from smaller cloud providers and regions outside the U.S. NVIDIA's improved GB200 chip is helping to meet the demand for AI servers. The OpenAI-SoftBank-Oracle "Stargate" project shows a shift towards bigger, project-based demand.

States' ESG fight offers lessons for powering AI's energy needs

AI is using a lot of energy, straining the power grid. One lesson comes from states pushing back against ESG finance, which limited energy funding. A study found that when states pushed back, markets adjusted without raising costs. Similarly, states must secure their energy futures to meet AI's demands. Some states are fast-tracking natural gas permits and investing in nuclear energy.

Equinox tests AI for workout and nutrition advice in its app

Equinox is testing a new AI feature in its app to give workout and nutrition advice. The tool is currently only available for employees to test. Equinox believes its own training data makes the AI's recommendations better than standard AI models. The AI can also learn from user feedback to improve future suggestions. Equinox also uses AI to recommend classes and answer questions like "What time is my gym open?"

AI oversight lacking leading to data breach risks, says IBM

Companies using AI without proper oversight are facing increased security risks. IBM's 2025 Cyber Resilient Organization study found that 16% of data breaches involved AI tools. Many companies experiencing AI-related breaches lacked access controls and AI governance policies. The average cost of a data breach in the United States is now $10.22 million. Companies are improving breach management, partly by using AI to monitor networks.

Brookfield increases investment in AI

This article could not be summarized. The content is not available.

SentinelOne buys Prompt Security to boost AI tool visibility for MSSPs

SentinelOne has acquired Prompt Security to help managed security service providers (MSSPs) track AI tool usage. This deal will give MSSPs real-time visibility and policy enforcement for AI tools. It helps them protect against data leaks and misuse without slowing down innovation. The platform integrates with SentinelOne’s Singularity Platform, enriching data with AI context. Prompt Security also detects Shadow AI, tools used without IT approval, and works with major AI models like OpenAI and Google.

Sources

AI Artificial Intelligence Copyright Data Mining AI Training Intellectual Property Australia Voice Actors Biometric Data AI Regulation Text and Data Mining Exception Indigenous Cultural Property Job Security AI and Jobs GPU AI Hardware AI Software Cloud Infrastructure Energy Consumption ESG AI in Fitness AI in Nutrition Data Breach Cybersecurity AI Governance MSSP Managed Security Service Providers Shadow AI AI Tool Visibility Prompt Security SentinelOne OXCapsule OXPython Y Combinator OpenAI Google NVIDIA GB200 Stargate project