What OpenAI’s New High-Level Reasoning Model Means for Legal Tech (Legal Tech News) "Another season, another OpenAI model. On Sept. 12, the artificial intelligence developer announced a new generative AI model series called OpenAI o1, unofficially and colloquially known as Strawberry, that improves its accuracy by reasoning first before responding to a problem or prompt.... 'Even if the gains are most pronounced in math and coding, don’t confuse that for the fact that there will still be a lot of gains in law and other sort of just general enterprise applications,' Pablo Arredondo, vice president of CoCounsel at Thomson Reuters, told Legaltech News." See also: The New Followup to ChatGPT Is Scarily Good at Deception (Vox) "To its credit, OpenAI had Strawberry tested by evaluators from outside the company. One of them, an organization called Apollo Research, specifically looked for evidence of scheming. And, well, they found it.... Basically, the AI figured out that if it has any hope of being deployed, it needs to present itself more like a hippie than like a business tycoon."  

OpenAI Pushes Prompt-Hacking Defense to Deflect Copyright Claims (Bloomberg) "OpenAI Inc. is making manipulation of ChatGPT central to its defense against a wave of copyright lawsuits from publishers and authors—accusing some of 'prompt hacking' and seeking vast troves of documents to learn exactly what questions they fed into the artificial intelligence tool to yield the outputs underlying their claims."  

RealPage Antitrust Suit Shows Importance of Vetting AI Tools (Bloomberg) "The civil antitrust suit brought in August by the Department of Justice and eight state attorneys general against property management technology company RealPage Inc. and several landlords using the software is yet another example of federal antitrust regulators taking an aggressive enforcement approach on new artificial intelligence technologies.... In light of the DOJ’s aggressive approach, companies using pricing algorithms should review how those algorithms operate, document all procompetitive benefits, and assess whether they are in fact indemnified or insured against an investigation."  

Deepfakes in Legal Proceedings: A Strategic Framework for Collaborative Solutions (Legal Tech News) "Deepfake technology poses a serious challenge to establishing and rebutting the authenticity of digital exhibits in legal proceedings.... To meet baseline duties of competence, lawyers must be prepared to detect and address deepfakes, to support a claim that audio or video evidence is fake, or to prove audio and video evidence is authentic."  

How to Use Geofence Warrants in a Constitutional Manner (Law360) Robert Frommer: "Geofence warrants are powerful tools that let law enforcement identify devices located at a specific location and time based on data users send to Google LLC and other tech companies. But left unchecked, they threaten to empower police to invade the security of millions of Americans. Thankfully, there is a way that geofence warrants can be used in a constitutional manner, if only courts would take it."  

How AI-Generated Memes Are Changing the 2024 Election (NPR) "Last week, Donald Trump posted on Truth Social an artificial intelligence-generated image of Taylor Swift in an Uncle Sam outfit, falsely claiming she had endorsed him. The image is clearly fake and was accompanied by other depictions, some also apparently AI-generated, of young women in T-shirts reading 'Swifties for Trump.' See also: Election Officials Say Candidates Can’t Mislead with AI Ads (Bloomberg).  

Why We Need An AI Safety Hotline (MIT Technology Review) Kevin Frazier: "In the past couple of years, regulators have been caught off guard again and again as tech companies compete to launch ever more advanced AI models.... In theory, external whistleblower protections could play a valuable role in the detection of AI risks. These could protect employees fired for disclosing corporate actions, and they could help make up for inadequate internal reporting mechanisms."  

Smart Cars at Heightened Risk of Attack, IT Security Firm Says (Bloomberg) "Automakers should do more to increase the safety of vehicles whose software can be updated over the air, according to a cybersecurity expert, because current technology leaves cars vulnerable to being manipulated without an owner’s knowledge.... There have been several instances where cybersecurity experts successfully sent commands to a vehicle remotely over the internet using an unauthorized account, according to Liz James, a consultant at IT security firm NCC Group, whose clients include some European automakers."  

23andMe Inks $30M Data Breach Deal with 6.4M Users in MDL (Law360) "Personal genomics company 23andMe has reached a $30 million settlement to resolve multidistrict class action litigation on behalf of more than 6 million customers whose personal data was stolen and in some cases leaked onto the dark web, according to a California federal court filing Thursday. Plaintiffs representing a putative class of about 6.4 million customers laid out the terms of the proposed deal in a motion for preliminary approval that touted the agreement as 'an outstanding result,' particularly in the face of 23andMe Inc.'s 'dire financial condition.'"  

San Francisco Sheriff’s GPS Tracking Program Divides 9th Cir. (Bloomberg) "A challenge to the San Francisco sheriff’s policy of sharing the GPS location data of criminal defendants released from custody ahead of their trials appeared to divide a three-judge US appeals court panel at oral arguments Tuesday. The city is seeking reversal of a district court injunction blocking that part of the sheriff’s pretrial electronic monitoring program that allows the sheriff to share location data with other law enforcement agencies for separate criminal investigations without a warrant."  

How to Tell If What You're Reading Was Written By AI (Lifehacker) "When you ask ChatGPT to write you something, the AI breaks down your question and identifies what it 'thinks' are the most important elements in your query. It then 'predicts' what the right sequence of words would be to answer your request, based on its understanding of the relationship between words.... That said, no matter how you coax the AI into responding, it is beholden to its training, and there will likely be signs such a piece of text was generated by an LLM. Here are some things to look out for."  

Dismissing All Your Notifications? 5 Ways to Make Them Effective (PC Magazine) "Do your notifications and reminders work for you? Do they get the job done? If not, you should make them more effective. Don't worry; most of what you need to do only needs to be done once. You're not going to hear a bunch of ongoing tasks that you need to do. We are in set-it-and-forget-it territory! So get ready for your life to get better with these little tweaks."  

TikTok Just Had the Most Important Two Hours of Its Life (CNN) "Who really controls TikTok’s magical algorithm — the US-based company that runs the app or its Chinese parent, ByteDance? That’s the question that bedeviled a trio of federal judges on Monday charged with deciding whether to allow the implementation of a law that could ultimately result in TikTok being banned for all Americans."  

OpenAI Hits 1 Million Paid Users for Business Version of ChatGPT (Bloomberg) "OpenAI now has more than 1 million paid users for the corporate versions of ChatGPT — a sign of growing demand from businesses for its chatbot despite significant competition. San Francisco-based OpenAI said Thursday that the figure includes the total number of people signed up to use its ChatGPT Team and Enterprise services, which are aimed at companies, as well as people at universities using its ChatGPT Edu product."