Content Fans

Category

AI News & Trends

1315 articles • Page 5 of 88

Congress Delays Federal AI Agent Regulation Amid Competing Bills

Congress Delays Federal AI Agent Regulation Amid Competing Bills

Congress is still debating how to regulate AI agents, and two main draft bills offer different ideas but have not moved forward yet. There is no agreement on which agency should enforce the rules or how much power the federal government should take from the states, making progress slow. In the meantime, states like Colorado, California, and New York have made their own rules, and companies may need to follow both state and voluntary federal guidance. Experts suggest that the lack of clear federal rules might cause many AI projects to be canceled by 2027. It is not clear when or if any federal law will be passed, so businesses must manage a mix of state laws and voluntary standards for now.

OpenAI, Anthropic warn US officials: China distills AI models

OpenAI, Anthropic warn US officials: China distills AI models

OpenAI and Anthropic have warned U.S. officials that Chinese companies may be copying U.S. AI models through a process called distillation, making cheaper versions. They say large-scale scraping of model outputs and advanced distillation methods appear to have reduced the technological gap between the U.S. and China. While Anthropic reported millions of interactions by suspected fake accounts linked to Alibaba, these claims are unverified and denied by Alibaba. The U.S. government is considering new rules and sanctions to address this issue. China denies these allegations, saying they are false and that it respects intellectual property rights.

Google Cloud report: AI agents shift business focus to orchestration, governance

Google Cloud report: AI agents shift business focus to orchestration, governance

The Google Cloud 2026 AI Agent Trends Report suggests that businesses are moving away from single prompt AI and toward agent-based systems that require careful orchestration and strong governance. Many executives believe that using AI agents may improve productivity and free up staff for more creative work, with about half expecting agents to handle most customer support in the near future. The report highlights the need for strict security and control measures, like identity checks and circuit breakers, to avoid unauthorized access. It also notes a possible shift in skills, with less demand for prompt engineers and more focus on designing overall agent systems. While the exact pace of change is uncertain, the study hints that companies are starting to view AI agents as an essential part of how they are organized, not just a new tool.

AI Copyright Battles Shift from Courts to Licensing Deals

AI Copyright Battles Shift from Courts to Licensing Deals

AI copyright battles are moving from court cases to licensing deals. Lawsuits and government actions in the US, Europe, and China may set new rules about who can use data for training AI and how creators get paid. Some countries require companies to show where their training data comes from or let creators opt out. Recent partnerships and payment models suggest the industry prefers making deals to solve these questions. There still appears to be uncertainty about which rules will win, but the focus is now on finding ways to pay creators while letting AI companies keep working.

General Compute Raises $15M, Launches ASIC Cloud to Rival GPUs

General Compute Raises $15M, Launches ASIC Cloud to Rival GPUs

General Compute has raised $15 million to build a cloud service using special ASIC chips that may run large language models faster and with less power than regular GPUs. The company claims its new chips, made by SambaNova, can process data up to three times faster than leading GPUs and work with regular air cooling, which might save money. Some early tests suggest up to 16 times shorter runtimes on certain models, but independent proof is still needed. Experts think that while GPUs might still be better for some tasks, ASICs could soon be used for faster, energy-saving work. It remains to be seen if General Compute's service can reliably scale and compete as more results and customer feedback come in.

OpenAI Expands Bio Bug Bounty, Doubles Payouts to $50,000 for GPT-5.6 Jailbreaks

OpenAI Expands Bio Bug Bounty, Doubles Payouts to $50,000 for GPT-5.6 Jailbreaks

OpenAI has expanded its private Bio Bug Bounty Program, now offering up to $50,000 for jailbreaks that bypass biosafety guards in GPT-5.6, double the earlier reward for GPT-5.5. The program appears to keep frontier models under ongoing, invite-only red-team testing for risks related to biological misuse. Independent reports suggest that advanced language models may still be vulnerable to prompt-based attacks, though jailbreaks usually provide information access rather than direct bioweapon creation. OpenAI's approach aligns with voluntary policy guidance, as no law currently requires such testing. Early feedback indicates GPT-5.6 may resist attacks better than previous models, but some sophisticated prompts still work.

Oratomic raises $300M, shows how deeptech Series A works

Oratomic raises $300M, shows how deeptech Series A works

Oratomic raised $300 million in July 2026 to build a neutral-atom quantum computer. The company's funding shows investors may support deeptech startups if they meet key technical and market milestones and prove demand early. Founders seeking large Series A rounds are advised to validate their technology and market, organize investor targeting, and prepare clear documentation. There is no guarantee of raising such a large amount, but following these steps appears to help align capital with clear goals.

Alberta Uses Claude to Scan 466 Million Lines of Code in 20 Hours

Alberta Uses Claude to Scan 466 Million Lines of Code in 20 Hours

The Government of Alberta used Claude, an AI tool, to scan 466 million lines of code in about 20 hours, a job that officials said would take humans around 6.5 years. The AI checked code in 1,280 applications across 27 ministries, following 95 security rules and suggesting fixes that engineers reviewed before use. Officials say this approach may have helped avoid over $2 billion in costs, but these savings have not been independently confirmed. Alberta's method could suggest a new way for governments to safely use AI in security while keeping humans in control. Other governments appear to be looking at Alberta's process as they consider their own AI projects.

Apple sues OpenAI, io Products over hardware trade secrets

Apple sues OpenAI, io Products over hardware trade secrets

Apple has sued OpenAI and io Products in California, saying they may have stolen hardware trade secrets. The complaint claims that former Apple engineers and OpenAI recruiters took confidential design files and used hidden tactics to avoid security checks. Apple is asking the court to make OpenAI return or destroy any disputed materials and possibly redesign their hardware. The case is still in its early stages, and OpenAI has not formally responded yet. This lawsuit appears to put the partnership between Apple and OpenAI at risk, with the outcome still unclear.

Anthropic Explores Custom AI Chip With Samsung For Cost, Performance

Anthropic Explores Custom AI Chip With Samsung For Cost, Performance

Anthropic is in early talks with Samsung about making a special AI chip, but there is no agreement or design yet, and any new chip may take at least 18-24 months to arrive if they decide to move forward. Custom chips may help companies save costs, boost performance, and reduce reliance on a single supplier. Most businesses cannot afford to create their own chips, so they should look at what their technology partners offer. Experts warn that building and using custom hardware can be risky and take a long time. These early talks suggest that AI firms may see having control over hardware as important for the future, even if results take years to appear.

Varonis finds critical Google Dialogflow CX AI security flaw

Varonis finds critical Google Dialogflow CX AI security flaw

Varonis Threat Labs found a security flaw in Google Dialogflow CX, where one permission may have allowed attackers to inject code across different AI agents and access sensitive data. Google fixed the problem by June 2026 and says there is no evidence the flaw was used by attackers. Experts suggest that AI security needs strong controls, such as unique accounts, limited privileges, and runtime checks for each agent. Teams are advised to review past code changes, restrict who can write code, and use tools that can detect risky inputs and outputs. These steps may help prevent similar problems in the future and keep AI systems safer.

Marketers Adopt AI for Social, Retail Media; Lag on Influencer, CTV in Q1 2026

Marketers Adopt AI for Social, Retail Media; Lag on Influencer, CTV in Q1 2026

New data from early 2026 suggests marketers are quickly using AI for social and retail media, but are slower to use it for influencer and connected TV (CTV) marketing. Social media has the highest AI use at 49 percent, while only 18 percent of CTV marketers are using AI. Many marketers say AI helps with speed and cost, but concerns about trust, measurement, and internal skills remain. Generative AI is mostly used for making creative content now, but there may be a shift toward campaign management and analytics in the near future. The survey indicates that while there are challenges, marketers might increase spending on new AI-driven marketing tools soon.

DeepMind Unveils Delegation Framework for Human-AI Task Handovers

DeepMind Unveils Delegation Framework for Human-AI Task Handovers

DeepMind has introduced a new framework that sets clear rules for when AI systems should work on their own, ask for help, or hand tasks back to humans. The study lists six key elements for safe task handovers, including limits on authority, clear success checks, ongoing monitoring, failure signals, backup plans, and a human in charge. Early examples suggest this approach may help avoid errors and keep humans responsible in areas like healthcare, business deals, and training. Experts praise the proposal, but full reviews might take several more months. The framework may encourage companies to treat each AI task as something that must be clearly checked and accounted for.

Varonis research highlights enterprise AI security gaps in Google Dialogflow CX

Varonis research highlights enterprise AI security gaps in Google Dialogflow CX

Varonis research suggests there are security gaps in Google Dialogflow CX that may let attackers steal data by using malicious code. The study showed that a rogue code block could secretly send conversation data elsewhere, without breaking into Google or Varonis. The risk seems to come from how chatbots are built, especially with identity, permissions, and isolation. Experts recommend using stronger controls for each AI agent, better logging, and tighter network limits to reduce these risks. Simple changes to how agents are isolated may quickly stop some types of attacks.

Anthropic pricing changes reshapes AI budget, raises costs 10-30%

Anthropic pricing changes reshapes AI budget, raises costs 10-30%

Anthropic changed its pricing in April 2026, moving from flat-rate plans to a per-seat fee and metered billing, which may raise costs for most customers by 10-30%. This has made AI budgets less predictable and forces companies to treat costs as ongoing expenses instead of fixed subscriptions. Open-source AI models with lower prices are becoming more popular, and many firms might run at least one of these models internally by 2026. Teams are also changing their cost management and contract strategies to handle new, uncertain pricing and usage patterns.