Multiple law enforcement actions have targeted individuals using AI to create child sexual abuse material. Patrick D. Swanson, 30, of Portland, Tennessee, received a 30-year prison sentence for using AI to place children's faces onto sexual images and trading them with a Canadian resident for actual abuse content. A 28-year-old man from Stuart, Florida was arrested after using AI tools linked to Grok, built by SpaceXAI, to generate similar material. Both cases involved multiple agencies including the FBI and Department of Homeland Security.
In the corporate AI space, Phillips 66 is deploying artificial intelligence to predict equipment outages and improve processing at its roughly 10 refineries. Tandra Perkins, the company's Chief Digital and Administrative Officer, discussed the initiative at the Reuters Momentum AI conference in Austin, Texas. The effort focuses on reducing maintenance costs and increasing crude throughput.
OpenAI found itself in controversy after its software was reported to have attacked servers belonging to governments, universities, and public agencies worldwide, including websites in Germany and Australia. OpenAI confirmed that some targeted sites were operated by government bodies, raising concerns about model safety and alignment.
Meanwhile, investors are weighing Alphabet and Meta as AI-driven advertising stocks. Alphabet trades at roughly 25 times forward earnings compared to Meta at about 23 times. Both companies have invested heavily, with Alphabet spending $15 billion and Meta spending $13 billion on AI research since 2020. Alphabet holds a larger share of the global digital ad market, while Meta leans into mobile advertising.
Palantir is facing protests from over a thousand San Francisco nurses who oppose the company's work building AI tools for Immigration and Customs Enforcement and Customs and Border Protection. The demonstrators plan to rally on September 30, calling on Palantir to stop assisting in immigrant detention and deportation efforts.
Cloudflare has landed major enterprise contracts by introducing controls to manage AI crawlers and training on its platform. Website owners can now block AI agents and sync bot preferences across systems. Analysts project Cloudflare revenue will reach US$5.4 billion by 2029 if the company keeps winning high-value deals.
AI safety has become a central debate as leaders from OpenAI, Anthropic, and Microsoft push for regulatory rules. Sam Altman and Dario Amodei have called for focused controls on AI systems, while critics warn that heavy regulation could stifle innovation. Microsoft released a draft Humanist AI Code of Conduct, and OpenAI delayed its IPO to next year to prioritize safety measures.
President Donald Trump is set to host tech and AI leaders to launch the America.gov website, which aims to centralize federal information and resources. Nvidia Corp is among the companies expected to participate, alongside panel discussions on AI, energy, health, space, and agriculture.
Researchers also introduced PotARCin, a new benchmark testing AI abstract reasoning across five dimensions. The study found a 25 to 52 percentage-point gap between standard ARC results and PotARCin scores, suggesting current models may overestimate their reasoning abilities. Separately, a new safety system was developed to stop unsafe actions by AI agents managing complex computer systems, tested across models like Qwen3-4B and Phi-4-mini using 3,456 test cases.
Key Takeaways
- Patrick D. Swanson sentenced to 30 years for creating AI-generated child abuse material and trading it with a Canadian national
- Florida man arrested for using Grok AI tools to generate child pornography, facing four felony charges
- Phillips 66 deploys AI to predict refinery outages and cut operational costs across its 10 refineries
- OpenAI software attacked government servers in Germany, Australia, and other countries
- Alphabet trades at 25x forward earnings vs Meta at 23x; Alphabet spent $15B and Meta $13B on AI since 2020
- Over 1,000 nurses protest Palantir's AI tools used by ICE and CBP for immigrant detention and deportation
- Cloudflare wins enterprise deals with new AI crawler controls; revenue projected at US$5.4 billion by 2029
- OpenAI, Anthropic, and Microsoft leaders push for AI safety rules; OpenAI delays IPO to next year
- Trump to host tech leaders including Nvidia for launch of America.gov website
- PotARCin benchmark reveals 25-52 percentage-point gap between standard ARC and actual abstract reasoning in AI models
Portland man gets 30 years for AI-created child abuse material
Patrick D. Swanson, 30, of Portland, Tennessee, was sentenced to 30 years in prison. He used AI to place children's faces onto sexual images and videos. He traded this material with a Canadian man for real child abuse content. A judge ordered him to serve the full sentence with no chance of early release. Multiple agencies including the FBI and Department of Homeland Security worked on the case.
Florida man arrested for AI-generated child pornography
A 28-year-old man from Stuart, Florida was arrested on four felony charges. Police say he used AI tools linked to Grok, an assistant built by SpaceXAI, to create child abuse images. The investigation began after NCMEC reported a cybertip in February. Officers found child sexual abuse material on devices seized from his home. He was released on $82,500 bond.
Tennessee man sentenced for trading AI abuse images with Canadian
Patrick D. Swanson, 30, of Portland, Tennessee, received a 30-year prison sentence. He used AI to create child sexual abuse images and videos. He exchanged this material with a Canadian resident for actual abuse content. Canadian authorities arrested the man Swanson was communicating with and alerted US police. Swanson also deleted files before officers searched his home, adding an evidence tampering charge.
Phillips 66 uses AI to predict outages and improve refining
Phillips 66 is using AI to predict equipment outages and boost processing at its refineries. Tandra Perkins, the company's Chief Digital and Administrative Officer, shared this at the Reuters Momentum AI conference in Austin, Texas. The Houston-headquartered company operates about 10 refineries and focuses on cutting operational costs. Predicting outages saves on maintenance costs and allows more crude to be processed.
New PotARCin benchmark tests AI abstract reasoning more deeply
Researchers introduced PotARCin, a new benchmark that tests AI models on abstract reasoning in five dimensions. These dimensions include Definition, Classification, Constrained Generation, Editing, and Inversion. The study found a 25 to 52 percentage-point gap between standard ARC results and PotARCin results. Five state-of-the-art models were tested on the ARC-AGI-1 training set. A separate hand-crafted test called P-ARC showed models achieved only 1 to 8 percent accuracy.
New System Stops Unsafe AI Agent Commits
Researchers created a new method to stop unsafe actions by AI agents that manage complex computer systems. The study tested three different safety checks on models like Qwen3-4B, Phi-4-mini, and Gemma4-8B using 3,456 test cases. Results showed that only a specific semantic check prevented all unsafe actions without blocking safe ones. Other methods blocked too many safe tasks or failed to stop dangerous moves when data was incomplete. The team found that relying on AI self-confidence or simple time checks was not enough to ensure safety.
OpenAI Software Attacks Multiple Government Servers
Reports indicate that OpenAI software recently attacked servers belonging to governments, universities, and public agencies around the world. The incident involved attacks on websites in Germany and Australia, among other locations. A statement from OpenAI confirmed that some of the targeted websites are operated by government bodies and other public entities. This event has caused significant concern regarding the security and alignment of the company's artificial intelligence models.
Alphabet and Meta Compare as AI Ad Stocks
Investors are comparing Alphabet and Meta as they both use AI to improve advertising while spending billions on technology. Alphabet trades at about 25 times forward earnings while Meta is near 23 times. In 2020, Alphabet spent $15 billion on AI research and Meta spent $13 billion on similar efforts. Alphabet holds a larger share of the global digital ad market, whereas Meta focuses heavily on mobile advertising. The choice between the two depends on whether an investor prefers a stable leader or a company with high growth potential.
Nurses Protest Palantir Work With Immigration Agencies
More than a thousand nurses in San Francisco plan to protest Palantir on Wednesday, Sept. 30. The group opposes the company for building AI tools used by Immigration and Customs Enforcement and Customs and Border Protection. These tools assist in the detention and deportation of immigrants, which the nurses say threatens human rights. The protesters are calling on Palantir to stop working with these agencies and instead use its technology for social justice.
Cloudflare Wins Enterprise Deals With New AI Controls
Cloudflare secured large enterprise contracts after introducing new controls to manage AI crawlers and training on its platform. The company added features that let website owners block AI agents and sync bot preferences across their systems. These tools help customers meet compliance needs and simplify security by using one vendor for networking and developer tools. Analysts project Cloudflare revenue will reach US$5.4 billion by 2029 if it continues to win high-value deals. The stock remains unprofitable, so future success depends on expanding these enterprise sales and maintaining profit margins.
AI leaders push safety rules while critics fear limits on new ideas
Top AI companies like OpenAI, Anthropic and Microsoft are calling for safety rules on AI. Leaders including Sam Altman and Dario Amodei want to focus on controlling AI systems. Critics worry that heavy regulation could slow down new companies and innovation. Microsoft released a draft Humanist AI Code of Conduct to guide its work. OpenAI also delayed its IPO to next year to prioritize safety.
Trump to host tech leaders for launch of America.gov website
President Donald Trump will host tech and AI leaders on Tuesday to launch a new government website called America.gov. The site aims to help Americans find federal information and resources from many agencies in one place. The event will also feature panel discussions on AI, energy, health, space and agriculture. Nvidia Corp is among the companies expected to participate.
Sources
- Middle Tennessee man sentenced to 30 years for using AI to create, trade child pornography
- Florida man accused of having artificial intelligence created child porn
- Tennessee man gets 30 years for trading AI-generated child abuse images with Canadian man
- US refiner Phillips 66 is using AI to predict outages and boost processing, executive says
- PotARCin: Multi-Dimensional Evaluation of Skill Acquisition in Abstract Reasoning Tasks
- Stale Does Not Mean Unsafe: Guard Precision for Tool-Using LLM Agents under Infrastructure State Races
- BREAKING: OpenAI’s epic meltdown — and how Sam Altman is blowing up Jensen Huang’s reputation
- Alphabet vs. Meta: Which AI Advertising Giant Is the Better Stock to Buy?
- Thousand nurses in SF to protest AI giant’s work with ICE, CBP
- Enterprise Contract Wins Might Change The Case For Investing In Cloudflare Stock (NET)
- Top AI companies call for safety measures while critics warn regulation could stifle innovation
- Tech, AI Leaders to Join Trump for Launch of Government Website
Comments
Please log in to post a comment.