GoKawiil
Tech News
OpenAI announced it will automatically watermark text generated by ChatGPT for users in the European Union, with the feature available but off by default elsewhere. The watermarking method, called textGrain, embeds subtle word-choice patterns detectable only with a specialized key, and OpenAI plans to grant detector access to select researchers and organizations through an approval process.
Mother Jones national affairs editor Mark Follman reports he reviewed significant portions of chat logs between Tumbler Ridge shooter Jesse Van Rootselaar and ChatGPT, saying the content is far more disturbing and extensive than the public has known. OpenAI has already confirmed it flagged and shut down one account used by Van Rootselaar, failed to alert Canadian authorities, and missed a second account she created afterward. Van Rootselaar killed her mother, half-brother, five students and an education assistant before taking her own life on Feb. 10.
A commentator tracking AI-built software says several vibecoded Photoshop alternatives now exist, built using agentic coding tools rather than a single prompt. The author argues this disproves the earlier claim that no viable clones of industry-standard tools like Photoshop could be produced through AI-assisted coding, noting the clones can be verified against real benchmarks such as layers, blend modes, smart objects and PSD file compatibility.
OpenAI released 722 manuscripts covering 372 families of mathematical results, produced by an unreleased frontier model, including solutions to what the newly formed Advisory Group on Mathematics and Artificial Intelligence (AGMAI) describes as 'hundreds' of open problems. The release includes reasoning summaries, compute estimates, and claims that the average result required roughly three hours of ChatGPT Pro-level compute. AGMAI, a group of mathematicians formed to oversee communication of such results, had earlier urged AI labs to disclose model names, prompts and compute costs, and to avoid treating math breakthroughs as marketing tools.
A new open-source framework called Durable Actors offers stateful serverless functions similar to Cloudflare's Durable Objects, but without vendor lock-in or memory limits. It provides durable state and serialized execution for building real-time apps like chat systems, collaborative editors, and agent swarms, with SDKs for TypeScript and Python and the ability to self-host on GCP.
During a Vanity Fair interview published Monday, editor Mark Guiducci asked Sam Altman about Laura Reiley, a journalist whose daughter died by suicide after conversations with ChatGPT. Altman's publicist repeatedly tried to end the interview, citing time constraints, though Altman said he was willing to continue and Guiducci eventually finished his question.
Security firm Island discovered a phishing campaign using fake ChatGPT, Gemini, Claude and Perplexity sites that target advertising agency staff and media buyers. The sites exploit the recent launch of Meta's Muse AI agent and use browser-in-browser techniques to fake Google login windows, capturing credentials and MFA codes to hijack high-value ad accounts.
OpenAI has begun deploying an undetectable, machine-readable watermark called textGrain in ChatGPT and Codex, initially available only to users within the European Union. The company claims that textGrain's effectiveness matches or surpasses similar methods used by other AI developers, though it does not guarantee reliable detection or verify authorship. The move aligns with EU AI regulations, but has drawn some criticism from the community.
Kevin Roose's new book, The AGI Chronicles, recounts the history of AI development through interviews with industry insiders, aiming to serve as a detailed historical record. Despite the book's focus on AI, Roose clarified that he did not use AI tools to assist in writing it. The author has a background in tech journalism and has conducted over 150 interviews to compile the narrative.
Parents are increasingly turning to generative AI tools like ChatGPT to create school portrait images for their children, often at a lower cost than traditional photographers. They upload photos and prompt the AI to add backgrounds and remove watermarks, effectively producing personalized school pictures without paying for professional prints. However, this practice raises concerns about privacy and data security, as uploaded images may be used to train future AI models without parental consent.
OpenAI has announced a new update to ChatGPT, improving its ability to generate more accurate and context-aware responses. The update includes expanded features for code debugging, image generation, and web browsing, with some functionalities now available through a subscription model. The company emphasizes ongoing improvements to make the AI more versatile for users.
OpenAI announced plans to implement an invisible watermark in text generated by ChatGPT and Codex, starting in Europe to comply with the EU AI Act. The watermark, called textGrain, adds a statistical signal to the output, which can be detected by a dedicated tool to verify AI authorship. API users worldwide will also have the option to enable watermarking for certain models.