As generative AI technology, like ChatGPT, advances, concerns about its misuse have intensified. OpenAI’s annual report highlights various abuses, revealing that some originate from countries like China. The report details ten cases where AI was exploited for misinformation, including accounts generating misleading social media posts and engaging in cyber activities, such as password cracking, linked to geopolitical interests. OpenAI emphasizes that understanding these threats helps refine their defenses. Other abusive uses were traced to actors from Russia, Iran, and other regions, indicating a widespread challenge. As more sophisticated AI applications emerge, such as text-to-video and text-to-speech, the potential for misuse grows. While developers implement safety measures, the creativity of malicious actors complicates the landscape, with a lack of robust federal oversight in the U.S. adding to the urgency of addressing these challenges.

Source link

Magistrate Judge Ona Wang swiftly granted the New York Times’ request regarding copyright concerns linked to ChatGPT, suggesting users might erase chats to bypass paywalls, hindering evidence collection for the plaintiffs. OpenAI is appealing this decision, arguing it constitutes an overreach by the Times. The company emphasizes its commitment to user trust and privacy through an FAQ that outlines data retention policies, clarifying that while deleted chats can be accessed under legal obligations, they won’t be automatically shared with the Times. Instead, this data is secured and can only be accessed by a small, authorized legal and security team. OpenAI reassures users amidst the ongoing lawsuit, uncertain about how long retention will last, but highlighting that API business customers under Zero Data Retention agreements are unaffected. The situation has prompted concerns among users about potentially switching to competitor services until clarity is restored.

Source link

The article presents a detailed coding tutorial for integrating SerpAPI with Google Gemini-1.5-Flash, focusing on advanced analytics. It covers essential steps, starting from setting up the APIs to executing complex queries that leverage Gemini’s capabilities for enhanced data analysis. The tutorial emphasizes best practices for efficient coding and maximizing the performance of the SerpAPI in conjunction with Gemini. Examples of practical applications and code snippets are provided to help users understand the integration process better. Additionally, it discusses potential challenges and offers troubleshooting tips. By the end of the tutorial, readers are equipped with the knowledge to effectively utilize both SerpAPI and Google Gemini, enabling them to perform advanced data analysis and improve their analytics projects.

Source link

Apple’s WWDC Keynote is set for Monday at 10 am PDT, where expectations are high, but certain unveilings, like the “personal” Siri, have been delayed. This feature, promoted in an ad featuring Bella Ramsey, is unlikely to debut until mid-to-late 2026. Along with this, advanced capabilities for Siri to interact with on-screen content and app controls have also been postponed. The hoped-for LLM Siri, which would operate similarly to generative AI apps like ChatGPT, is estimated to be two years away. Additionally, AI enhancements for the Calendar and Health apps will not be discussed this year, with significant updates earmarked for iOS 27. On a more immediate note, the iOS 26 beta will be released post-Keynote, and users are advised to back up their data before installation, as initial versions are often buggy.

Source link

Gemini, the cryptocurrency exchange founded by the Winklevoss twins, has submitted a draft registration statement to the U.S. Securities and Exchange Commission (SEC) for a proposed initial public offering (IPO). This move aims to enhance transparency and compliance within the crypto industry while capitalizing on the growing interest in digital assets. The filing indicates Gemini’s commitment to regulatory standards, as it seeks to provide a robust platform for trading cryptocurrencies. The IPO is part of a broader trend where cryptocurrency companies are transitioning towards traditional financial structures. While the details regarding the number of shares and pricing have not yet been disclosed, the announcement highlights Gemini’s strategic direction and ambition in the evolving financial landscape. As the market continues to mature, this IPO could set a precedent for other crypto firms considering similar steps towards public listing.

Source link

Google’s Gemini app is introducing a new feature called “scheduled actions,” enhancing its functionality for everyday tasks. This update allows users with a Google AI Pro, Ultra subscription, or supported Google Workspace plans to automate specific tasks at designated times or on a recurring basis. Notably, it works across platforms, benefiting iPhone users with the app installed. Gemini can now proactively manage routine tasks, such as sending daily calendar summaries, weekly blog ideas, or sports updates, as well as handling one-time requests. Users can easily manage scheduled actions within the app’s settings by specifying tasks and times. This update aligns with Google’s goal of transforming Gemini from a reactive chatbot into a more helpful daily assistant, making it a practical enhancement for organization and task management. The feature is currently available for eligible users on both Android and iOS without needing a new app download.

Source link

AI is revolutionizing video creation, moving beyond just deepfakes and filters to allow users to generate high-quality videos purely through text prompts. Two leading platforms, OpenAI’s Sora and Google’s Veo, exemplify this innovation. Sora, integrated into ChatGPT, allows users to create 10-second videos (720p) with a Plus subscription and 20-second videos (1080p) with a Pro subscription. Veo promises cinematic quality, enabling 60-second videos with advanced features like camera movements and lighting changes through simple commands. To use Veo, users can subscribe to Google’s Gemini plans, which vary in pricing and features. A recent test comparing a golden retriever puppy video on both platforms revealed Sora’s strength in visual detail, while Veo excelled in lifelike motion. This shift towards effortless video production democratizes storytelling for marketers, creators, educators, and enthusiasts, hinting at a bright future in AI-driven content creation.

Source link

Employees at the Department of Government Efficiency (DOGE) used a flawed AI model to assess contracts at the Department of Veterans Affairs (VA), leading to the cancellation of hundreds of contracts worth millions. Under a tight 30-day deadline to respond to a directive from President Trump, engineer Sahil Lavingia rushed to develop the AI, which aimed to “munch” contracts not directly supporting patient care. However, the AI’s methodology was severely flawed, as it only analyzed the first 10,000 characters of contracts, lacked precise definitions for critical terms, and was built on an outdated model. Consequently, many essential contracts were flagged inaccurately. Experts noted that the AI couldn’t comprehend the specifics required for such evaluations. Though Lavingia acknowledged mistakes were made, he claimed all flagged contracts were vetted by others. Initially, 875 contracts were set for cancellation, but this was later revised to about 585 after pushback from veteran advocates and stakeholders.

Source link