AI News

Follow the latest AI and automation news

19 Jun 20264 Muharram 1448 AH

The US banned Anthropic’s Fable 5 release, but the numbers don’t seem to care

The US government forced Anthropic to withdraw its new models, Fable 5 and Mythos 5, due to national security concerns. This decision followed Amazon researchers discovering a way to bypass security barriers in the Fable 5 model. Cybersecurity researchers expressed alarm over this action, signing an open letter warning of the risks, noting similar vulnerabilities exist in other models. Anthropic confirmed that these vulnerabilities are not unique to Fable 5. This development reflects the growing tension between AI innovation and national security requirements, raising questions about future regulations for this technology.

Business

19 Jun 20264 Muharram 1448 AH

Billionaire Ambani wants AI in every call, app, and home

Reliance has announced the integration of AI technologies into its telecom services, impacting over 500 million users. This shift comes as demand for smart and reliable services in the telecom sector increases. The company aims to enhance user experience by utilizing AI for data analysis and personalized service offerings, reflecting a global trend towards digital transformation. This innovation strengthens Reliance's competitive edge in the growing telecom market.

Business

19 Jun 20264 Muharram 1448 AH

Norway to Impose Near-Total Ban on AI in Primary Schools

Norwegian Prime Minister Jonas Gahr Støre announced a near-total ban on the use of AI tools in primary schools, with restrictions on their use for older students. This decision follows a decline in educational test results, and the government has prohibited smartphones in schools since 2024. The new standards will be implemented starting next academic year, prohibiting AI use for students in grades one to seven, while allowing lower secondary students to use it cautiously under teacher supervision. Støre emphasizes the importance of teaching children reading, writing, and arithmetic.

💡

SAP and Google Cloud deploy agentic commerce architecture

How can AI enhance customer experience in retail?

Models

A startup claims it broke through a bottleneck that’s holding back LLMs

19 Jun 20264 Muharram 1448 AH

Models

19 Jun 20264 Muharram 1448 AH

A startup claims it broke through a bottleneck that’s holding back LLMs

Miami-based AI startup Subquadratic has introduced a new LLM called SubQ, claiming it is faster, cheaper, and more energy-efficient than existing models. SubQ reportedly processes up to 12 times more text simultaneously, enabling it to handle data-intensive tasks like analyzing hundreds of documents or entire codebases. Despite initial skepticism, Subquadratic has released independent evaluation results from Appen, which lend credibility to its claims. Co-founder Alex Whedon noted that sharing these benchmarks earlier would have mitigated much of the doubt surrounding the model.

Infrastructure

Global Optical Chip Makers Accelerate Capacity Expansion as AI Data Center Demand Surges

19 Jun 20264 Muharram 1448 AH

Infrastructure

19 Jun 20264 Muharram 1448 AH

Global Optical Chip Makers Accelerate Capacity Expansion as AI Data Center Demand Surges

The global optical chip industry is experiencing unprecedented expansions and strategic agreements. On June 16, Coherent signed a letter of intent for up to $50 million in funding from the U.S. Department of Commerce to expand its Texas facility, doubling manufacturing space and quadrupling production capacity. NVIDIA's founder, Jensen Huang, attended the groundbreaking, announcing a $2 billion strategic investment in Coherent. Huang emphasized that AI relies on compute, but scaling is limited by connectivity, highlighting the factory's critical role in meeting rising demand.

Models

OpenAI researchers show small doses of "beneficial trait" training make AI models broadly safer and harder to manipulate

19 Jun 20264 Muharram 1448 AH

Models

19 Jun 20264 Muharram 1448 AH

OpenAI researchers show small doses of "beneficial trait" training make AI models broadly safer and harder to manipulate

OpenAI researchers demonstrated that reinforcement learning focused on desirable traits like truthfulness and corrigibility is effective across various domains. Training on health data also enhanced deception detection, with the model outperforming on 44 out of 53 benchmarks. This approach contrasts with Anthropic's constitution-based method. This advancement suggests that reinforcement learning can make AI models safer and less susceptible to manipulation, marking a significant step towards improving the reliability of intelligent systems.

Agents

e2e-assure introduces Cumulo, the U.K.’s only sovereign, AI-driven, zero-day SOC platform to secure IT and OT environments

19 Jun 20264 Muharram 1448 AH

Agents

19 Jun 20264 Muharram 1448 AH

e2e-assure introduces Cumulo, the U.K.’s only sovereign, AI-driven, zero-day SOC platform to secure IT and OT environments

e2e-assure has launched Cumulo, the U.K.'s only AI-first SOC platform aimed at enhancing cyber defenses. This initiative responds to GCHQ Director Anne Keast-Butler's call for a new national cyber defense capability integrating agentic AI. Cumulo features native AI integration, allowing continuous context building as security data is generated. The platform introduces a zero-day SOC model, enabling immediate application of new threat intelligence as detection rules, thus mitigating risks from emerging threats.

Governance

Google appeals ruling that made it directly liable for AI-generated search overview content

19 Jun 20264 Muharram 1448 AH

Governance

19 Jun 20264 Muharram 1448 AH

Google appeals ruling that made it directly liable for AI-generated search overview content

Google is appealing a ruling from the Munich Regional Court that held it liable for inaccurate AI-generated search results. The court found that the AI incorrectly linked two local publishers to fraud schemes, a matter the court deemed serious despite Google's claim of "minor errors." This case highlights growing concerns over the accuracy of information provided by AI systems. The ruling sets a legal precedent, emphasizing the need for accountability in AI applications, which could influence how companies manage their AI technologies moving forward.

Business

Source: Elastic agrees to buy CRV-backed DeductiveAI for up to $85M

19 Jun 20264 Muharram 1448 AH

Business

19 Jun 20264 Muharram 1448 AH

Source: Elastic agrees to buy CRV-backed DeductiveAI for up to $85M

DeductiveAI, a startup founded three years ago, focuses on using artificial intelligence to discover and resolve software bugs. The company offers innovative solutions aimed at improving software quality and reducing the time spent on debugging. As reliance on software increases, tools like DeductiveAI become essential for enhancing efficiency and cutting costs. DeductiveAI is developing advanced AI algorithms that enable faster and more accurate code analysis and bug detection. This technology represents a significant shift in handling software errors, reducing the need for human intervention.