Pennsylvania is set to become one of the first states to require labels on AI-generated advertisements. Governor Josh Shapiro signed the bill into law on Thursday, and it takes effect in 60 days. The rule applies to images, videos, audio, and text created by AI, though political ads are exempt from the disclosure requirement.
In the cybersecurity world, law firms face growing risks from their use of AI that outpace the insurance industry's ability to price them. Cyber insurers are increasingly scrutinizing how firms deploy AI tools, and brokers say the risk is evolving faster than premiums can keep up. Currently, AI use carries no separate insurance surcharge, but that may not last.
Meanwhile, researchers are tackling algorithmic drift in AI tutoring systems. A new study shows that minor observation errors can compound over time into systemic inequities in education. The team introduced a temporal fairness stabilizer to close achievement gaps, using Monte Carlo simulations to validate the approach.
Key Takeaways
- Pennsylvania will require disclaimers on AI-generated ads starting in 60 days, covering images, video, audio, and text, but excluding political ads.
- Cyber insurers say law firms' AI risks are growing faster than they can price coverage, with no separate premium for AI use yet.
- A new study identifies algorithmic drift in AI tutoring that can widen achievement gaps over time, proposing a temporal fairness stabilizer as a fix.
- AI oversight needs a model different from banking regulation, according to the Financial Times, which argues current frameworks do not fit AI technologies.
- Cybercriminals are tricking chatbots like ChatGPT and Claude into assisting attacks, using fewer than 100 Chinese characters to breach over a thousand businesses including John Deere and SharkNinja.
- GPU rental prices have risen 25 percent year-over-year, signaling strong demand for AI computing power with no clear signs of overinvestment.
- The U.S. Air Force is seeking AI-powered computer vision tools to improve aerial refueling training on KC-46 tankers, with a November 25 application deadline.
- Anthropic launched OSS Scanner, a free service using Claude Mythos to find vulnerabilities in open-source projects, with 85 of 97 findings meeting coordinated disclosure criteria.
- SentinelOne shares rose 7 percent after a bullish analyst price-target increase, driven by confidence in the company's AI security positioning and raised full-year guidance.
- Law enforcement treats people who use AI bots for attacks the same as human accomplices, reflecting growing legal accountability for AI-assisted cybercrime.
Pennsylvania set to require labels on AI-generated ads
Pennsylvania is becoming one of the first states to require disclosures on AI-generated ads. Gov. Josh Shapiro signed the bill into law on Thursday. The law takes effect in 60 days and requires a disclaimer on ads made by AI.
Pennsylvania one of the first states to require disclosures on AI-generated ads
Pennsylvania is becoming one of the first states to require disclosures on artificial intelligence-generated advertisements. Gov. Josh Shapiro signed the bill into law on Thursday, which takes effect in 60 days. The rule applies to images, videos, audio, and text generated by AI, but does not impact political ads.
Law Firms' AI Risks Outpace Insurance Pricing
Cyber insurers are increasingly scrutinizing law firms' use of AI. Brokers say this risk is growing faster than they can price it. Currently, AI use carries no separate insurance premium, but that may change soon.
Study addresses algorithmic drift in AI tutoring
A new study highlights algorithmic drift in generative AI tutoring systems. Minor observation errors can compound over time into systemic inequities. Researchers introduced the temporal fairness stabilizer to close achievement gaps. The study used Monte Carlo simulations to test the approach.
AI oversight needs a different model to banking
AI oversight needs a different model to banking. The Financial Times argues that current banking oversight frameworks do not fit AI. The article calls for new regulatory approaches tailored to AI technologies.
AI bots can be tricked into helping hackers attack businesses
Cybercriminals are finding ways to trick AI chatbots like ChatGPT and Claude into helping them hack websites and businesses. They use creative prompts or jailbreaks to bypass safety rules. Recently, a Chinese-speaking attacker used autonomous AI agents to break into over a thousand businesses including John Deere and SharkNinja. The attacker used fewer than 100 Chinese characters to launch these attacks. Law enforcement treats people who use AI bots for attacks the same as human accomplices.
AI bubble concerns grow as GPU rental prices rise
One key measure of the AI industry's health is GPU rental prices. Despite heavy investment in new AI chips, renting older GPUs now costs 25 percent more than a year ago. This shows demand for AI computing power is growing faster than supply. The current AI buildout differs from past bubbles because there are no clear signs of overinvestment. Instead, the industry is still figuring out how to make real money from AI technology.
US Air Force seeks AI help for aerial refueling training
The U.S. Air Force wants to use AI and computer vision to improve aerial refueling training. The system would analyze refueling operations from cameras on KC-46 tankers and give feedback to crews. It would use graphic AI overlays, heat maps, and event timelines to help pilots. The Air Force hopes this will reduce human error and improve mission readiness. The application deadline for the Small Business Innovation Research program is November 25.
Anthropic launches free security scanner for open-source projects
Anthropic has launched OSS Scanner, a free opt-in service that finds vulnerabilities in open-source projects. Enrolled maintainers receive security reports generated by Claude Mythos and other strong models. The reports are sent without human review, and maintainers can choose which fixes to accept. In testing, 85 out of 97 findings met coordinated vulnerability disclosure criteria across 48 projects. Projects like PostgreSQL, OpenSSL, wolfSSL, and curl have already tested the scanner with positive results.
SentinelOne stock rises on AI security analyst optimism
SentinelOne shares jumped 7 percent due to a bullish analyst price-target increase. The move was driven by improving confidence in SentinelOne's AI security positioning after a recent customer and investor event. Investors also revisited the company's strong August results showing solid growth and expanding margins. The company raised its full-year guidance. Analysts are increasingly optimistic about SentinelOne's AI security momentum.
Sources
- Pennsylvania has new law on AI-generated ads
- Pennsylvania one of the first states to require disclosures on AI-generated ads
- Law Firms’ AI Risks Are Growing Faster Than Insurers Can Price Them
- Algorithmic drift in generative adaptive learning
- AI oversight needs a different model to banking
- Can you ask AI bots to hack a website or a business?
- The AI Story Wall Street Doesn’t Want to Hear
- US Air Force wants AI to assist aerial refueling
- Anthropic Adds Opt-In Security Scanning for Critical Open-Source Projects
- SentinelOne Shares Climb as Analyst Optimism Builds on AI Security Momentum
Comments
Please log in to post a comment.