Google DeepMind, OpenAI, and Anthropic just agreed to grant the UK government early access to their AI models for safety evaluations. Prime Minister Rishi Sunak announced the deal alongside a £100M taskforce and an upcoming global summit, aiming to position the UK as the world capital of AI guardrails.
Google just introduced SAIF, a comprehensive security framework designed to protect AI systems from unique threats like prompt injection, model theft, and data poisoning. Drawing from its zero-trust experience, the tech giant outlines six core principles to help enterprises secure deployments by default. Why it matters: As generative AI scales, standardizing guardrails will be critical to keeping rogue prompts from breaking enterprise systems.
Ron DeSantis’ presidential campaign just released an attack ad against Donald Trump featuring three AI-generated deepfake images of Trump embracing Anthony Fauci. Forensic experts flagged telltale flaws like mangled White House signage and airbrushed textures, highlighting how political campaigns are now actively blurring reality with synthetic media.
University of Kansas researchers just trained a machine learning classifier that accurately spots ChatGPT-generated science papers over 99% of the time, dropping to 92% at the paragraph level. The tool analyzes stylistic differences like vocabulary richness and punctuation habits that AI often misses. Why it matters: Academic publishers are desperate to catch AI slop, though small sample sizes and human light edits could still slip through.
OpenAI was just hit with what appears to be the first-ever defamation lawsuit after ChatGPT fabricated accusations against a radio host. The chatbot falsely claimed Mark Walters embezzled over $5 million from a non-profit while summarizing a legal document for a journalist. While internet platforms usually enjoy Section 230 immunity, an AI generating defamatory claims from scratch could rewrite the legal playbook.
Apple CEO Tim Cook just revealed he personally uses ChatGPT and confirmed Apple is looking at the tech closely. While excited about its promise, Cook warned about bias and misinformation, arguing that because AI moves too fast for regulation, companies must step up and police themselves.
The EU just urged tech giants including Google and Facebook to immediately start labeling AI-generated text and images to combat online disinformation. Regulators want tags noticeable enough that users scrolling by instantly know "the robot is talking," while also warning Twitter that defying incoming digital laws could trigger fines up to 6% of global turnover.
Top AI researchers and tech CEOs just signed a stark, one-sentence warning comparing the extinction risk of artificial intelligence to pandemics and nuclear war. Organized by the Center for AI Safety, signatories include OpenAI's Sam Altman, Google DeepMind's Demis Hassabis, and Turing Award winner Geoffrey Hinton. Why it matters: The concise statement shows unprecedented alignment across rival labs to force catastrophic AI risks into mainstream policymaking.
Japanese officials just clarified that developers can train AI models on copyrighted works without individual licenses, so long as the goal is data analysis rather than enjoying the creative expression. The exception extends even to commercial systems, though generated outputs that infringe on existing works can still cross legal lines. Why it matters: Japan is removing legal uncertainty to build a competitive domestic AI ecosystem, putting speed ahead of its massive anime and manga industries' compensation worries.
A New York attorney used ChatGPT for legal research and cited six entirely hallucinated cases in a court filing. When ordered to produce the opinions, he had ChatGPT generate bogus rulings and submitted the screenshots. The federal judge has ordered a sanctions hearing for everyone involved.
OpenAI just launched a grant program offering ten $100,000 grants to fund experiments in setting up democratic processes for deciding AI rules. Winning teams must engage at least 500 participants and open-source their code and reports by October 20, 2023. Why it matters: OpenAI believes decisions about AI behavior should reflect broad public oversight rather than being dictated by a single company.
Sam Altman just warned that OpenAI could cease operating in the EU if it can't comply with the upcoming AI Act. Altman pushed back on classifying general-purpose models like GPT-4 as "high risk," noting there are technical limits to compliance.
A fake, AI-generated image showing a massive blast near the Pentagon just went viral on Twitter, briefly knocking the US stock market down 0.26% before it recovered. Spread by high-profile accounts including Russian state media RT, the hoax was quickly debunked by the local Arlington Fire Department, while eagle-eyed users spotted warped fences and classic AI artifacts.
OpenAI just published a framework for governing future superintelligence, arguing systems exceeding human expert levels could arrive within a decade. The team proposes an international body modeled after the IAEA to inspect frontier models and track compute, while keeping lower-tier models and open source free from heavy red tape.
Apple just banned employees from using ChatGPT and GitHub Copilot over concerns that confidential code or project data could be leaked or sucked into model training. Even with chat history turned off, OpenAI keeps conversations for 30 days. Apple now joins the likes of JPMorgan and Amazon in locking down generative AI on the job.
OpenAI is now going after developers who put "GPT" in their domain names. A developer revealed on Hacker News that brand protection firm BrandShield sent a takedown notice over an unused domain containing "GPT". OpenAI warns that putting model names directly in site titles or URLs risks confusing consumers, meaning projects can only use phrases like "Powered by GPT".
OpenAI CEO Sam Altman just called for strict government regulation in a Senate hearing. He proposed a licensing and registration system for powerful AI models, alongside mandatory safety evaluations before release. It is a proactive push by OpenAI to shape the emerging legal playbook.
Microsoft CEO Satya Nadella just pushed back against Elon Musk's claim that Microsoft controls OpenAI through its multibillion-dollar investment. Nadella clarified that Microsoft only holds a non-controlling interest and commercial partnership, while a non-profit board steers the startup. He added that smaller players still have plenty of room to take on tech giants.
Elon Musk just claimed in a CNBC interview that he is the reason OpenAI exists, arguing he came up with the name, recruited key talent, and invested roughly $50 million. Musk also warned that the startup no longer prioritizes AI safety and is effectively controlled by Microsoft, an assertion Microsoft CEO Satya Nadella strongly denied.
OpenAI CEO Sam Altman just testified before a Senate subcommittee for the first time, urging lawmakers to create dedicated regulatory agencies and licensing systems for advanced AI models. Senators pledged not to repeat the past mistakes made with social media, praising Altman's cooperative stance as night and day compared to other Big Tech executives.
