Top AI, Cloud & Security News - This Week in AI with NexaQuanta

This Week at NexaQuanta

Welcome to the latest edition of the NexaQuanta newsletter. This week in AI, we bring you the most impactful developments shaping the world of artificial intelligence, data infrastructure, and digital innovation.

Whether you’re an executive, developer, researcher, or just AI-curious, our updates will keep you informed and ahead of the curve.

In this issue, we cover IBM’s latest warning on rising AI-related data breaches, along with their powerful new watsonx data integration platform.

We take a close look at Google’s AI-powered search launch in the UK and Microsoft’s massive earnings report driven by cloud and AI investments.

Plus, OpenAI’s new “Study Mode” aims to reshape student learning, and AWS’s central AI and cloud infrastructure updates round out a transformative month in tech.

IBM Report Reveals AI-Driven Breaches on the Rise, U.S. Breach Costs Hit Record $10.22M

AI Security Falling Behind Adoption

IBM’s newly released Cost of a Data Breach Report 2025 has uncovered a growing risk linked to artificial intelligence. While organizations are rapidly adopting AI, many are neglecting essential security measures.

The report shows that 13% of organizations experienced AI-related breaches. Among them, 97% lacked proper access controls for their AI systems.

The consequences were severe. Sixty percent of these AI incidents led to compromised data, and 31% disrupted operations.

The report points out a key issue: AI systems are being launched without adequate oversight or governance.

Shadow AI and Smarter Attacks

One in five organizations reported breaches due to shadow AI — unauthorized or unmanaged AI tools. These incidents resulted in higher breach costs and more compromised personal and business data than average.

At the same time, attackers are starting to use AI tools themselves. About 16% of breaches included AI-powered attacks, mainly through phishing and deepfakes.

Cost and Recovery

Globally, the average cost of a data breach has decreased to $4.44 million. However, the U.S. saw an increase, with the average breach cost climbing to $10.22 million.

Healthcare remains the costliest sector for breaches, with an average impact of $7.42 million and the longest time to recover.

Despite the rising costs, only 49% of breached organizations plan to invest in stronger security, a drop from 63% the year before.

AI Security Investments Still Lacking

The report also highlights that many organizations lack AI governance policies. Only 34% of those with a policy conduct regular audits. This leaves sensitive data and AI models open to manipulation.

However, companies that used automation and AI in their security response were able to save an average of $1.9 million in breach costs and reduce their response time by 80 days.

Operational and Financial Impact

Most breached organizations faced significant operational disruption. Recovery took over 100 days on average.

Nearly half of them said they would raise prices to deal with the financial impact—about one-third of those planned price increases of 15% or more.

Click here to read more about this news.

IBM Launches watsonx.data Integration to Unlock Unstructured Data for AI

A Unified Platform for Structured and Unstructured Pipelines

IBM has introduced watsonx.data Integration, a powerful solution that allows organizations to build pipelines for both structured and unstructured data using a single control plane.

Launched on June 11, 2025, the platform is designed to meet the growing demand for data integration in enterprise AI workflows.

The solution includes built-in observability, support for batch, streaming, and real-time replication pipelines.

With its unstructured data integration (UDI) capabilities, it simplifies the process of handling complex formats like PDFs, PPTs, images, and more.

Key Capabilities of watsonx.data Integration

Prebuilt connectors
Enables ingestion of diverse unstructured data types at scale while preserving metadata and access controls.
Drag-and-drop transformation
Includes IBM Research–backed operators for tasks like text cleaning, PII masking, hate and abuse detection, and quality filtering—no coding required.
Vectorization and embeddings
Converts unstructured files into vectors for AI tasks like RAG, classification, and semantic search.
Scalable architecture
Efficiently processes petabytes of data, compressing large documents for rapid performance.
Dynamic updates
Automatically tracks document changes and updates only the modified content, saving time and resources.
Secure access control
Maintains document-level permissions throughout the pipeline for compliance and trust.

Built on Open-Source Innovation

IBM built this solution from real-world experience, including training its Granite foundation models.

The system incorporates technologies like the Data Prep Kit (DPK) and Model Factory (DMF)—now open-sourced via the Linux Foundation.

It also integrates Watson Document Understanding and Docling, which offer high-speed and accurate entity extraction.

Support for Modern AI Workflows

The platform works with vector stores such as Milvus and managed services. It’s also being integrated with orchestration tools like Langchain, enabling developers to build full-stack AI pipelines with enterprise-grade governance.

Transforming Enterprise Data into AI-Ready Insights

IBM’s watsonx.data integration transforms previously underutilized enterprise data into valuable insights. With most enterprise data being unstructured—and only 1% currently used in generative AI—the platform bridges a significant gap in data readiness.

This launch positions IBM as a leader in enabling organizations to harness unstructured content for real-time decision-making, content generation, and intelligent automation.

To read more about this news, click here.

Google Rolls Out AI-Powered Search Mode in the UK

Fewer Links, More Answers — And Big Implications for the Web

Google has launched its AI Mode search feature in the UK, marking a significant shift in how users interact with the world’s most-used search engine.

Instead of displaying a list of clickable links, the new tool provides answers in a conversational tone generated by Google’s Gemini AI platform.

AI Mode has already been introduced in the US and India. The rollout in the UK is expected to be completed in the coming days.

For now, it remains optional and can be accessed via a tab or directly in the search bar.

Changing How People Search

The update is designed to respond to more natural, complex user queries. For example, instead of typing “clean carpet stain,” users might now ask, “I spilled coffee on my Berber carpet, I’m looking for a cleaner that is pet friendly.” AI Mode is meant to handle these detailed questions more effectively.

Google says this shift represents a new way of seeking information, one that goes beyond traditional keyword-based search.

Concerns for Publishers and Businesses

The feature is drawing concern from publishers and web-based businesses. Since AI responses contain fewer direct links, traffic to websites—especially news outlets—could drop.

The Daily Mail reports that its traffic from Google has already fallen by around 50% on both desktop and mobile since the earlier AI Overview feature was introduced.

A Pew Research study also found that users clicked on links just once in every 100 searches when AI summaries were present. Google disputes this data, calling the study method flawed.

No Clear Revenue Model Yet

Currently, it’s unclear how advertising will work within AI Mode. Google has not confirmed whether companies will be able to pay for placement inside AI-generated responses.

This uncertainty adds to the concerns of businesses that depend on search engine visibility for customer acquisition.

Environmental and Regulatory Factors

AI Mode is not available in the EU due to existing regulations. There are also growing concerns about the environmental impact of AI.

The technology requires large-scale data centers, which consume significant energy and clean water. Google says it is working to maintain sustainability as it expands its AI capabilities.

A Fundamental Shift in Search Behavior

While AI Mode won’t replace traditional Google Search, it introduces a new direction—one where users stay on Google’s platform longer and may bypass external websites entirely.

This development could reshape the digital economy, from news distribution to advertising strategies.

Click this link to read more details about this news.

Microsoft Reports $27.2B Profit, Validates Massive AI and Cloud Investments

Strong Financials Reflect Payoff of AI Strategy

Microsoft has announced a $27.2 billion quarterly profit, signaling that its major bets on artificial intelligence and cloud infrastructure are beginning to deliver strong financial returns.

The company’s total sales reached $76.4 billion for the April–June quarter, up 18% from the same period last year.

The results exceeded Wall Street forecasts and pushed Microsoft’s stock price up more than 8% in after-hours trading.

Massive AI and Cloud Spending

Over the past year, Microsoft has spent $88 billion building data centers to support AI and cloud computing.

These investments, while initially raising investor concerns, now appear to be driving both growth and profit.

The company’s full-year profit, ending June 30, reached over $101 billion—a milestone that highlights the scale and profitability of its AI-led strategy.

More Spending Ahead

Microsoft told investors to expect continued expansion of its data center infrastructure in the upcoming quarter.

The company is confident that demand for AI services will keep rising, and it plans to support that demand with even more capacity and innovation.

Want to read more about this? Click here!

OpenAI Unveils ‘Study Mode’ to Support Smarter Learning in Students

New ChatGPT feature aims to enhance understanding, not just deliver answers

OpenAI has launched a new feature in ChatGPT called “study mode”, designed to help students solve problems through guided steps rather than shortcuts.

This feature is available for Free, Plus, Pro, and Team users, and will be integrated into ChatGPT Edu in the coming weeks.

Helping Students Think, Not Just Cheat

The rapid adoption of AI tools like ChatGPT in education has raised concerns about students using them to bypass critical thinking.

OpenAI’s new study mode is a direct response. Instead of giving straight answers, it asks guiding questions, encouraging users to think through problems, prepare for tests, and grasp new material more thoroughly.

Leah Belsky, OpenAI’s VP of Education, explained that when ChatGPT is used as a tutor, academic performance improves. But if used solely to fetch answers, learning suffers.

Built With Educators and Students

Study mode was developed with input from teachers, researchers, and students through OpenAI’s ChatGPT Lab. In a demo video, OpenAI showed how students interact with the tool by going through multiple steps and summarizing the answer in their own words.

Reimagining Education

This development follows recent comments by OpenAI CEO Sam Altman, who suggested that college might not be relevant in the future.

Despite this, OpenAI continues to work closely with academic institutions, having already launched ChatGPT Edu for universities last year.

Study mode represents a step forward in using AI to support—not replace—real learning.

Click here to read more details about the news.

AWS Summit NYC 2025: Major AI & Cloud Infrastructure Announcements

Amazon Web Services outlines its next steps in AI scalability, customization, and infrastructure

At the AWS Summit New York 2025, held on July 16, AWS made major announcements around AI agent development, cloud scalability, and infrastructure improvements.

The keynote by Swami Sivasubramanian, VP of Agentic AI at AWS, emphasized production-ready AI and streamlined cloud operations.

Key Highlights from the Event:

AgentCore (Preview): A new framework for deploying and managing AI agents with built-in enterprise features like memory management and identity control.
Customizing Amazon Nova via SageMaker: Full lifecycle customization—pre-training, fine-tuning, and alignment—now possible for Amazon Nova foundation models.
AWS AI League: A hands-on learning initiative for enterprises to develop AI skills in prompt engineering and model tuning.
Expanded Free Tier: AWS is offering new users $200 in service credits—$100 at sign-up and another $100 for using select services like EC2 and Bedrock.
TwelveLabs on Bedrock: Advanced video understanding models are now available, supporting tasks like video search and scene summarization.
Amazon S3 Enhancements:
- Metadata Updates: Improved visibility with SQL-ready inventory and journaling.
- Vector Storage (Preview): Optimized for AI tools with scalable querying and lower storage costs.
SageMaker Upgrades:
- QuickSight integration for visualizations
- Support for unstructured S3 data
- Auto-onboarding from Lakehouse data sources
Amazon EventBridge Logging: Detailed logging improves observability for event-driven applications.
ECS Blue/Green Deployments: Safer container rollouts with rollback capabilities now built in.
EKS Scaling: Clusters can now support up to 100,000 nodes—scaling to over 1.6 million Trainium chips or 800,000 NVIDIA GPUs for large AI/ML workloads.

These updates reinforce AWS’s strategic focus on making AI not only scalable but deeply integrated across enterprise workflows and cloud ecosystems.