NerdNews

Microsoft made three major AI announcements, including a new AI engineering group, pay-as-you-go agents for Copilot Chat, and bundling AI features into Microsoft 365. The company is prioritizing AI development and aims to automate tasks and provide AI-powered tools to businesses and consumers.

Microsoft has introduced its AI assistant Copilot to Microsoft 365 Personal and Family subscriptions for an extra $3 per month. The move has been met with mixed reactions from users, with some finding it helpful and others frustrated by its presence in Word and other apps. The feature can be disabled, but the process is not straightforward.

Microsoft's AI red team found that AI models amplify existing security risks and create new ones, making it impossible to secure them completely. The team identified eight lessons, including the need to understand AI capabilities, consider security implications, and use automation to cover more of the risk landscape.

Nvidia and Cisco have introduced new tools to help secure AI systems. Nvidia's NeMo Guardrails include three new microservices to prevent AI agents from being hijacked or producing unwanted results. Cisco's AI Defense offers model validation and discovery tools to help security teams identify and mitigate AI-related risks. These tools aim to address common problems in AI development and deployment, such as data leaks, prompt injection, and hallucinations.

The article discusses how AI risks in 2025 will arise from misuse, such as over-reliance on AI, intentional misuse like non-consensual deepfakes, and using AI for decisions without proper understanding. Examples include lawyers using ChatGPT to generate erroneous court briefings and companies exploiting public confusion to sell dubious products labeled as 'AI'. The authors argue that mitigating these risks is a task for companies, governments, and society, and that focusing on superintelligence is a distraction from the real issues.

Lior Barak and Gaëlle Seret discuss the importance of proactive, long-term, business-focused strategies for data engineering teams, emphasizing collaboration and alignment with stakeholders to drive impactful data products.

AI will change data engineering by building smarter data pipelines, reducing data mapping, and enhancing business intelligence. Data engineers must acquire new skills to apply AI models, manage third-party services, and focus on data strategy.

Instagram has added new features to its Reels platform, making it easier for users to discover and engage with content. The update comes as TikTok is set to be banned in the US, potentially driving users to alternative platforms like Instagram Reels. The new features include a button that shows the content liked or commented on by friends and followers, and a 'reply bar' that allows users to respond to content directly.

A network of over 200 websites is using AI-generated content and plagiarized articles to mimic legitimate sports news outlets, deceiving readers and siphoning off ad revenue. The sites often feature programmatic ads from major companies, and their use of AI tools allows them to replicate and scale their content quickly.

The article discusses the importance of securing non-human identities (NHIs) in a zero trust architecture. It highlights the risks associated with poorly managed NHIs, such as secrets sprawl and over-permissioned NHIs. The article provides guidance from NIST on securing NHIs, including eliminating long-lived credentials, enforcing least privilege, and using centralized secrets management. It also introduces GitGuardian as a solution to help organizations implement these security measures.

Nvidia is constructing a 30-megawatt research-and-development supercomputer in northern Israel, featuring hundreds of liquid-cooled Blackwell systems, at an estimated cost of $500M. The system will be used for developing next-gen datacenter technologies and is expected to start operations in the first half of 2025.

The US Supreme Court is set to hear TikTok's appeal against a nationwide ban. If the ban is upheld, it won't immediately remove the app from phones, but it will make it harder to use over time by preventing updates and limiting hosting services. Users may still find ways to access the app using VPNs or international services, but the experience may be degraded.

Nvidia announced Cosmos, a family of foundational AI models that can be used to train humanoid robots, industrial robots, and self-driving cars. Cosmos was trained on 20 million hours of real footage of humans and can generate realistic video footage of various scenarios, allowing robots to learn from them.

The Nintendo Switch 2 has been officially revealed with a larger display, magnetic Joy-Cons, and a sleeker design. It will be backwards compatible with Switch games and feature a new button on the Joy-Con. The console's specs, price, and launch games are still unknown, but more information will be revealed during a Nintendo Direct on April 2.

Google has reintroduced digital fingerprinting, a method of tracking users online, despite previous concerns over user privacy. This move allows advertisers to target users across devices, but has been criticized by the UK's Information Commissioner's Office as an unfair means of tracking users, reducing choice and control over personal information.

Google has added AI features, including Gemini, to its Workspace plans, increasing the base price from $12 to $14 per seat. Existing users who paid extra for Gemini will now pay less, while those who didn't use it will pay more. The changes will roll out immediately for new customers and on March 17 for existing accounts.

Nexos.ai, founded by Nord Security's Tomas Okmanas and Eimantas Sabaliauskas, aims to help enterprises deploy and manage AI models, focusing on large language models (LLMs). The platform provides visibility, security, and adaptability, allowing companies to access over 200 AI models through a simple API. With $8 million in funding, Nexos.ai plans to launch by the end of March, addressing the gap between AI pilot projects and production readiness.

The article highlights the top 5 CES 2025 announcements to look out for, including Honda's EVs, 4K and 5K OLED monitors, Lenovo Legion GO S and GO 2, Nvidia's new graphics cards, and a unique 'emotional-support AI robot'. These announcements are expected to showcase the latest advancements in technology, including AI, gaming, and automotive innovations.

The use of multi-cloud strategies is increasing, with 78% of organizations using two or more cloud providers, driven by the need for resilience and specialized services. However, security and compliance concerns persist, with 61% of organizations citing these as top concerns. The adoption of cloud security platforms and tools, such as Cloud Security Posture Management (CSPM) and Cloud-Native Application Protection Platforms (CNAPP), is on the rise to mitigate misconfigurations and compliance risks.

The article provides a comprehensive guide to protecting oneself from government surveillance, including using end-to-end encrypted messengers like Signal, encrypting devices, and being mindful of online anonymity and location data. It also discusses the importance of financial privacy and the potential risks of using burner phones. The guide is particularly relevant in the context of a new Trump administration, which has vowed to expand government surveillance.