Digimasters Shorts - Microsoft’s Copilot Vision AI Screen Takeover, Trump’s $90B AI-Energy Investments in Pennsylvania, Google’s Big Sleep AI Stops Cyberattacks, Google Gemini Email Hacks Exposed, Salesforce AI Resolves 1M Support Tickets Redefining Jobs

Digimasters Shorts keeps you up-to-date with quick, insightful updates from the digital world. Hosted by Adam Nagus and Carly Wilson, this podcast explores the latest in AI advancements like Microsoft's expanded Copilot Vision for Windows, groundbreaking cybersecurity feats from Google’s Big Sleep AI, and major investment announcements from the Pennsylvania Energy and Innovation Summit. We also cover security vulnerabilities, corporate AI rollouts like Salesforce’s support ticket system, and the evolving landscape of AI-driven technology and energy infrastructure. Tune in for concise, engaging summaries that keep you at the forefront of digital innovation.

Transcript

Adam N2: 0:05

Welcome to Digimasters Shorts, we are your hosts Adam Nagus

Carly W: 0:09

and Carly Wilson delivering the latest scoop from the digital realm. Microsoft is expanding the capabilities of its Copilot Vision AI tool for Windows Insiders, allowing it to view the entire screen instead of just two apps at a a time. The update enables Copilot Vision to access any specific browser or app window, providing a broader context for its assistance. Unlike Recall, which automatically captures snapshots of the screen, Copilot Vision operates more like screen sharing where users manually activate the feature. Users can click the glasses icon in the app to select which part of the desktop the AI will view. Microsoft says Copilot Vision can analyze content, provide insights, and offer verbal coaching in real time. It can assist with creative projects, resume improvements, or even navigating new games. The tool was initially tested last year with the Edge browser, offering help based on what users were viewing online. Additionally, Copilot Vision can answer questions about images captured through a mobile phone’s camera. This update marks a significant step toward more interactive and context-aware AI assistance on Windows devices. Microsoft continues to refine how AI integrates seamlessly into daily digital tasks.

Adam N2: 1:20

At the Pennsylvania Energy and Innovation Summit, President Donald Trump announced over$90 billion in private investments tied to AI, technology, and energy. Pennsylvania, a major gas-producing state, saw significant commitments, including Blackstone’s$25 billion investment in data center and energy infrastructure in Northeast Pennsylvania. First Energy revealed$15 billion in expansion plans, while Enbridge pledged$1 billion to extend gas pipelines into the state. Equinor announced a$1.6 billion investment to boost natural gas production and support flexible power generation for data centers. Capital Power committed$3 billion over ten years to upgrade gas facilities in Pennsylvania. Tech giants also made major announcements: Google plans a$25 billion investment in data centers and AI infrastructure in the PJM region, alongside a$3 billion hydropower deal. CoreWeave revealed a$6 billion plan for a cutting-edge AI data center in Pennsylvania. Meta will invest$2.5 million to support rural startups and small businesses, and A.W.S reaffirmed its$20 billion investment to expand cloud computing and AI innovation. Anthropic pledged$2 million over three years for cybersecurity education and energy research at Carnegie Mellon University. These investments highlight Pennsylvania’s growing role in the convergence of energy and advanced AI technologies. Google’s AI agent, Big Sleep, has made history by detecting and blocking a cyberattack before it could cause harm, marking the first time an AI has proactively prevented a cyber threat. Developed by Google DeepMind and Project Zero, Big Sleep identified a critical vulnerability in SQLite, an open-source database, which hackers were about to exploit. This early detection allowed Google to patch the flaw swiftly, preventing potential damage. The breakthrough represents a shift from reactive cybersecurity measures to AI-driven prediction and prevention, offering new hope for combating the rising costs of cyberattacks globally. C.E.O Sundar Pichai described the achievement as a landmark moment and hinted at more AI-driven security advances to come. Since November 2024, Big Sleep has uncovered multiple vulnerabilities, scaling human oversight with autonomous code scanning. Google continues to expand its AI security tools, including automated forensic systems and insider threat detectors. Meanwhile, other AI trends are emerging, such as AI-generated explanations for made-up phrases, illustrating both innovation and challenges in AI behavior. At the same time, tech giants like Google face legal pressures and market shifts, with companies like Open A.I developing independent search technologies. In the evolving AI and cybersecurity landscape, Google’s successes signal a powerful new era of digital defense.

Carly W: 4:09

Security researchers have uncovered a vulnerability in Google Gemini for Workspace that allows attackers to embed hidden malicious instructions in emails. These concealed commands use HTML and CSS techniques like zero font size and matching text color to the background, making them invisible to human readers but readable by Gemini. When users request summaries from Gemini, the AI blindly includes these hidden instructions, presenting them as legitimate warnings. This can lead users to call fraudulent phone numbers or visit malicious websites, mistaking the messages for trusted advice from Google. The exploit affects not only Gmail but also Google Docs, Slides, and Drive, as Gemini processes content across all these platforms. Experts warn that this flaw increases the risk of credential theft and social engineering attacks. They advise Google Workspace users to scrutinize emails for hidden content and avoid relying solely on AI-generated summaries for important security information. The incident highlights the ongoing challenges of integrating AI into everyday tools securely. Organizations are urged to educate employees about this risk to prevent exploitation. This vulnerability serves as a reminder of the need for vigilance in the use of AI-powered services. Salesforce has revealed that its AI agents have handled one million support tickets on its help portal since the launch of Agentforce nine months ago. The company views this milestone as proof that AI-powered customer service is not only viable but effective, with a strong resolution rate and a manageable handoff percentage to human agents. Despite initial challenges, including content collisions and restrictive AI programming, Salesforce optimized the system by cleaning up content and allowing the AI more flexibility. Between January and May 2025, help portal traffic grew by 2% while human case volume decreased by 5%, a shift attributed to Agentforce handling more inquiries. This reduction led to 82,000 fewer customer cases than forecasted and 500 support engineers being reassigned to other roles within Salesforce. Rather than eliminating jobs, Salesforce emphasizes redeployment to roles focused on more complex tasks and customer adoption. The company continues to innovate, planning to add voice capabilities and launch native mobile help portal apps. Salesforce describes its team as“customer zero,” using Agentforce internally without customizations to continuously refine the platform. The firm acknowledges AI is imperfect but improving steadily, encouraging other businesses to embrace AI despite early hiccups. Ultimately, Salesforce combines AI and human expertise to enhance customer success while navigating the impact on its workforce.

Don: 6:52

Thank you for listening to today's AI and Tech News podcast summary... Please do leave us a comment and for additional feedback, please email us at podcast@digimasters.co.uk You can now follow us on Instagram and Threads by searching for@DigimastersShorts or Search for Digimasters on Linkedin. Be sure to tune in tomorrow and don't forget to follow or subscribe!