Meta launches Llama 4 as Google Gemini scans photos

Key Takeaways

• Congressman Sam Liccardo supports a federal AI regulation framework, while Meta and OpenAI advocate for federal preemption to block state-level AI laws.
• The White House met with Anthropic, maker of the AI tool Claude Mythos, to discuss safety protocols for scaling AI, following reports of its advanced hacking capabilities.
• AGIBOT launched new embodied AI robots, including the humanoid AGIBOT A3, and foundation models for large-scale real-world deployment across industrial and service sectors.
• IBM is integrating AI, such as Retrieval Augmented Generation (RAG) and agents, to enhance mainframe operations, improve productivity, and automate tasks like support ticket generation.
• Schematik, an AI assistant, helps design hardware by generating code, and Anthropic has integrated a Bluetooth API for developers to build hardware interacting with its Claude AI.
• Google's 'Personal Intelligence' upgrade allows its Gemini AI to scan user photos for personalized image generation, an opt-in feature raising privacy considerations.
• AI is accelerating legal and financial services in Australia, with firms like Hicksons using it for document analysis and the superannuation sector exploring automated financial planning.
• Bugcrowd highlights increasing AI security risks, noting AI can find vulnerabilities faster than fixes, and warns about AI agents with broad permissions posing data breach threats.
• AI agents like OpenClaw, while offering automation, present significant security risks due to their extensive access to applications and data, potentially leading to loss of user control and data breaches.
• Users are expected to gain more control over their social media feeds, though specific implementation details are not yet available.

San Jose lawmaker at center of AI regulation debate

San Jose Congressman Sam Liccardo is involved in a national discussion about who should regulate artificial intelligence (AI). A group of child safety and tech watchdog organizations wants him to reject an endorsement from a pro-AI super PAC called Leading the Future. This PAC has ties to the Trump administration and companies working with U.S. Immigration and Customs Enforcement. Liccardo's office stated he supports a sensible federal AI regulation framework over a patchwork of state laws. Campaign finance records show neither Liccardo nor his supporting PAC have received money from Leading the Future.

Big Tech pushes to block state AI laws

Major technology companies are pushing for federal preemption to prevent states from creating their own artificial intelligence (AI) regulations. This means companies like Meta and OpenAI want to develop advanced AI without state-level safeguards. Critics argue that federal preemption undermines democracy and removes states' ability to protect their citizens from AI harms. They believe states should have the right to enact their own AI safeguards, as they are more responsive to local needs. Experts emphasize that supporting AI safeguards does not mean opposing AI development.

White House meets Anthropic amid AI hacking tool fears

The White House held a meeting with Anthropic, the company behind the AI tool Claude Mythos, which can reportedly perform advanced hacking tasks. This meeting occurred despite the White House previously labeling Anthropic a 'radical left, woke company.' Anthropic CEO Dario Amodei discussed collaboration and safety protocols for scaling AI technology with Treasury Secretary Scott Bessent and White House Chief of Staff Susie Wiles. The meeting also touched on the balance between AI innovation and safety. This comes after Anthropic faced a 'supply chain risk' label from the U.S. government, which a court partially agreed was retaliatory.

AI speeds up legal and financial services in Australia

Artificial intelligence (AI) is significantly speeding up how Australians interact with businesses, especially in the legal and financial sectors. Law firms like Hicksons are using AI to quickly scan and analyze large volumes of documents, saving time and helping junior lawyers learn faster. In insurance, Compare Club uses an AI digital assistant to help customers find the right health policies based on their needs. The superannuation sector is exploring AI to automate financial planning advice, guiding members on decisions about combining super, insurance, and investments. However, experts note Australia needs to increase its AI adoption to compete globally.

New control over social media feeds is coming

Users are expected to gain more control over their social media feeds soon. This development aims to give individuals greater influence over the content they see online. The specifics of how this control will be implemented are not detailed in the provided text. Further information is needed to understand the full impact of this change on user experience and platform algorithms.

AGIBOT launches new AI robots for real-world use

AGIBOT has introduced a new generation of embodied AI robots and foundation models designed for large-scale deployment in the real world. The company unveiled four new robotic platforms and AI models based on its 'One Robotic Body, Three Intelligences' architecture. These advancements aim to bridge the gap between advanced AI and practical productivity in industrial, commercial, and service settings. AGIBOT's new robots include the humanoid AGIBOT A3 for interactive environments and the AGIBOT G2 Air for human-machine collaboration, alongside the OmniHand 3 Ultra-T gripper.

IBM executive discusses AI enhancing mainframes

IBM Principal Product Manager Dan Wiegand explained how Artificial Intelligence (AI) is becoming essential for daily productivity and improving mainframe operations. He highlighted that AI, including techniques like Retrieval Augmented Generation (RAG) and agents, helps ground AI models with specific information and automates tasks. Wiegand emphasized that mainframes remain critical for businesses and that IBM is integrating AI to make them more accessible and efficient. RAG ensures AI responses are accurate and relevant to mainframe contexts, while agents automate tasks like opening support tickets.

Schematik AI helps build hardware, Anthropic integrates

Schematik, an AI assistant created by Samuel Beek, helps users design and build hardware devices by generating code for physical electronics. Beek developed Schematik after an AI-generated design for an electric door opener caused his house's fuses to blow. The tool aims to make hardware creation more accessible, similar to software coding. Anthropic has now integrated a Bluetooth API for developers to build hardware that interacts with its Claude AI. This move allows for more creative hardware projects, moving beyond software and images.

Google scans photos for AI image generation

Google has updated its services to allow AI image generation tools, like Gemini, to scan all user photos. This feature, part of Google's 'Personal Intelligence' upgrade, enables Gemini to use actual images of users and their loved ones in generated pictures. While this offers personalized AI results, it raises privacy concerns about Google accessing intimate personal moments. The feature is opt-in and currently rolling out in the U.S., with Google assuring users that privacy commitments remain unchanged. Users can adjust these settings at any time.

Bugcrowd highlights AI security risks and growth

Bugcrowd is focusing on the growing security risks associated with AI, noting that AI can find vulnerabilities faster than they can be fixed. The company emphasizes the need for rapid, risk-based prioritization to protect critical assets. Bugcrowd also warns about security threats from AI agents embedded in software, which often have broad permissions. They are promoting their crowdsourced approach to test AI systems and are expanding into sectors like financial services. Bugcrowd is also strengthening its presence in the public sector through partnerships and authorizations.

AI agents pose security risks despite convenience

The rise of AI agents, like OpenClaw, promises to automate tasks and save users time, but cybersecurity experts are concerned about the lurking security threats. These agents, powered by advanced AI, require access to numerous applications and data, increasing the risk of misuse. Experts warn that users may lose control over what these agents do, potentially exceeding set boundaries. While AI agents can improve efficiency, especially for remote workers and freelancers, they also present risks of data breaches and information disclosure, requiring careful consideration of security implications.

Sources

• San Jose lawmaker at center of AI regulation fight
• Big Tech wants to block state AI laws. Why it matters to you
• White House and Anthropic set aside court fight to meet amid fears over Mythos model
• How lawyers are using AI, and financial planners aren't far behind
• We're finally getting more control over social media feeds
• AGIBOT Unveils New Generation of Embodied AI Robots and Models, Accelerating Real-World Deployment of Physical AI
• IBM's Dan Wiegand on AI and Mainframe Augmentation
• Schematik Is ‘Cursor for Hardware.’ Anthropic Wants In
• Google Starts Scanning All Your Photos As New Update Goes Live
• Bugcrowd Leverages AI Security Push and Public-Sector Momentum in Latest Weekly Developments
• AI 'agent' fever comes with lurking security threats