Latest news with #Perplexity.ai
Time Magazine
06-05-2025
- Science
- Time Magazine
AI Won't Wait for Your Ethics
By Working at the intersection of work and technology, I hear the same concerns about artificial intelligence from women—particularly progressive, well-educated women—on repeat: 'I worry about AI's bias.' 'I have privacy concerns.' 'What about the environmental impact?' 'Is this cheating?' Yes, AI systems continue to reinforce existing societal biases — racial, gendered, and homophobic. Yes, there are real privacy risks, including deepfake technology that disproportionately targets women. And yes, the environmental impact is significant. GPT-4's training, for example, may have emitted up to 15 metric tons of CO₂—the same annual emissions as 938 Americans. These are real problems. But, while women name the issues, because the work needed to address them is so large, they disengage. Here's the irony: For women especially, the very values that drive us to challenge power may be making us more susceptible to it. While we debate, scrutinize, and hesitate, the systems we seek to reform move forward—unbothered and unchallenged. The result? AI is being shaped. Just not by us. The alternative: You don't need to abandon your values, you need to operationalize them. Here's how: If your concern is bias: Don't just critique from the sidelines—get in the room. AI learns from us. Our prompts, questions, and corrections shape what it becomes. If someone unlike you trains it, the bias stays baked in. Want to go further? Support or join teams building inclusive datasets. Learn how bias shows up in these systems so you can spot it and shift it. If your concern is privacy: Be intentional. Learn how your data is used. Many AI tools have trust centers where you can review their policies. Anthropic offers strong privacy protections. doesn't require a login. You also have the option to turn off model training. Prioritizing data privacy matters. Nonetheless, it's important to understand that you're also stepping back from influencing how these tools evolve. There's no perfect stance, only trade-offs. Still, using AI, even with training off, keeps you engaged and learning, which is better than checking out entirely. Choose tools that give you control. Push for better standards. Disengaging doesn't protect you. It just sidelines you. If your concern is the environmental impact: Apply the same scrutiny you would to food, fashion, or travel. Understand what energy a tool uses, and whether the company discloses emissions data. Models like DeepSeek are already challenging assumptions about energy use and creating pressure for more efficiency models. And not all AI use is equal. Generating text is far less taxing than creating video or high-res images. Fun fact: that 30-minute TikTok doomscroll before bed can use nearly 50 times more energy than a single ChatGPT prompt. Start with tools that publish transparency reports and use renewable energy—like Claude by Anthropic. Choose your tools and prompts thoughtfully. If you wonder 'Is this cheating?': I've never— not once —heard a man say that. Instead, they say it makes them more innovative. Using AI isn't inherently cheating. You can choose to use it to cut corners and offload your judgment and engagement. Or you can use it to expand your knowledge and deepen your thinking. When used well, AI becomes a collaborator. Use it to brainstorm, pressure-test your thinking, learn a language or some lesson about the world, or refine a messy first draft. It doesn't have to replace your voice—it can sharpen it. The real skill lies in applying judgment, strategy, and critical thinking. If you're still questioning whether AI is good for the world, well, it's complicated. But what's not complicated is this: the AI revolution is already underway. And without women shaping it, the outcome will be far worse.
Reuters
16-02-2025
- Business
- Reuters
From AI chatbots to social media: data identification and preservation in the digital age
February 12, 2025 - The advent of new software applications, such as AI chatbots like ChatGPT and third-party messaging applications like Slack, may help streamline chaotic areas of our lives, but their inclusion in e-discovery preservation workflows could do quite the opposite. New applications may pose challenges for collections in the course of discovery, but they more immediately pose challenges at the identification and preservation stages of the EDRM (Electronic Discovery Reference Model). As such, attorneys need to be aware of new applications their clients are using so that they can ensure data generated by this software is preserved appropriately and can counsel their clients on what to do if the application does not allow for the preservation of this data. Identifying new sources of data and the challenges they pose for preservation The new applications an average client is likely to use can be placed in one of three categories: AI chatbots, third party messaging applications, and social media platforms. Once you have identified one of these sources as containing relevant data, the Federal Rules demand that this data is preserved so that sanctions are avoided. The most common way that data is deleted from these applications is through default or selected settings. As such, the most integral step in data preservation for these applications is to understand what, if any, the default data retention period of a particular application is and if the period can be actively changed by the individual user for preservation purposes. Note that, as technology changes, so too do the retention policies associated with these applications. Be sure to consult each application's website to understand its latest privacy and retention policies. 1. AI chatbots An AI chatbot is a tool that uses artificial intelligence to answer questions and respond to prompts. Some of the most popular AI chatbots are ChatGPT and Microsoft Copilot; however, new chatbots are always emerging. For example, Jasper AI, and Google's Gemini have garnered the attention of technology enthusiasts. ChatGPT chats are saved unless actively deleted; however, users can opt to disable the automatic saving of chat histories. As such, ensuring that a client does not actively delete relevant histories and that automatic retention settings are not disabled will prevent relevant information from being deleted. Copilot chats, especially when governed by a larger organization, can be retained for a set period of time before deletion. In this instance, retention settings are set at an enterprise level for an organization, so it best for an attorney to understand what the retention period is and ensure automatic deletion is suspended if needed. prompts are saved as long as the account is active, and once an account is deleted, information is removed from servers within 30 days. Here, instructing a client to keep his/her account active and to not actively delete chat histories will ensure messages are preserved. If a deletion occurs within the 30-day window, corresponding directly with for preservation may be warranted. Jasper AI appears to have a short retention period for chat history so it is important to collect data as early as possible, to understand if there are any other means by which relevant chat history was saved if deleted (i.e., screenshots), and to document the steps taken to attempt preservation in case needed in court. Gemini prompts are saved for up to 72 hours, and the last 24 hours of conversation data is accessible within a user's activity log even if a particular activity function is disabled. Auto-delete settings are managed by the user with options to save prompts for anywhere from three to 36 months. As such, instructing your client to choose the longest auto-delete setting along with making early arrangements for collection will prevent spoliation. 2. Third party messaging applications Third party messaging applications allow individuals to do everything from chat and conduct video conferences to store shared files. Teams and Slack are rampant in the corporate market, WhatsApp and WeChat are popular in the foreign data market, and Signal, Telegram, and Element have also gained traction. These tools often have back-end enterprise settings that dictate how long messages are kept. Teams messages are stored according to an organization's retention policy, and policies can be set for specific teams or an entire organization. Understanding your client's back-end retention policies for these applications will ensure that you counsel your client appropriately about disabling auto-delete functions. Slack saves messages for different lengths of time depending on the user's plan and set retention schedules. For free accounts, data is kept up to one year, and for paid plans, data is kept for as long as the paid workspace exists by default. As such, it is important to understand under which type of account your client is operating so that you can appropriately counsel them on maintaining an active paid workspace or implementing an early collection so messages that would be deleted in the ordinary course are preserved. For WhatsApp and Element, users can set messages to disappear after a set amount of time, though messages can be kept for the lifetime of the account. For Telegram, Signal and WeChat, data is retained until an account is deleted or the user actively deletes the data. For these applications, attorneys should counsel their clients to choose appropriate retention settings when available and avoid deleting an account or relevant messages. 3. Social media accounts Social media has remained a staple of content production. The popular platforms of Facebook, Instagram and X are still thriving alongside Snapchat and TikTok, but other new platforms have emerged such as Bluesky, Threads, and Mastodon. Social media applications often offer lengthy data retention. Platforms like TikTok, X, Facebook, Instagram, Mastodon, Bluesky, and Threads do not delete your posted information unless the account itself is deleted or if the user actively deletes the data. Oftentimes, even if an account is deleted, there is a period during which the company may keep a user's data before permanent deletion. It is best to counsel your client to not actively delete any relevant data from these platforms and to keep his/her account active. If an account is deleted, it may be worth checking directly with the platform to see if that data has been retained and if it can be preserved. Snapchat data deletes 24 hours after viewing or seven days after viewing, depending on a user's setting; however, the Memories feature of Snapchat allows content to be kept for as long as the user would like. In this instance, it is best to counsel your client to suspend the use of these applications for relevant communications, work on an early collection of anything that may be preserved through a client's Memories, and document your process in the interim. Keeping up with the future of data identification and preservation The rapid evolution of software applications presents both opportunities and challenges in the realm of e-discovery. As new tools emerge, it is crucial for attorneys to stay informed about the technologies their clients use and their associated data retention policies. By understanding the default data retention settings and advising clients on best practices, legal professionals can ensure that relevant information is preserved appropriately.
