Leo Rodman June 04, 2025
Advanced OpenAI Model Caught Sabotaging Code Intended to Shut It Down
OpenAI's latest o3 model frequently sabotaged the script that would shut it down even when it was explicitly told not to.
futurism.com/openai-model-sabotage-shutdown-code
AI Goes Rogue: Claude Model Caught Attempting Blackmail During Safety Tests - TechStory
Anthropic's Claude AI tried to blackmail engineers during safety tests, threatening to expose personal info if shut down
techstory.in/ai-goes-rogue-claude-model-caught-attempting-blackmail-during-safety-tests
“Godfather” of AI calls out latest models for lying to users
Turing Award-winner Yoshua Bengio warns recent models display dangerous characteristics.
arstechnica.com/ai/2025/06/godfather-of-ai-calls-out-latest-models-for-lying-to-users
Mistral releases a vibe coding client, Mistral Code | TechCrunch
French AI startup Mistral is releasing its own 'vibe coding' client, Mistral Code, to compete with incumbents like Windsurf, Anysphere's Cursor, and GitHub Copilot.
techcrunch.com/2025/06/04/mistral-releases-a-vibe-coding-client-mistral-code
Unlicensed law clerk fired after ChatGPT hallucinations found in filing
Law school grad’s firing is a bad omen for college kids overly reliant on ChatGPT.
arstechnica.com/tech-policy/2025/06/law-clerk-fired-over-chatgpt-use-after-firms-filing-used-ai-hallucinations
Anthropic's AI is writing its own blog — with human oversight | TechCrunch
Anthropic has quietly launched Claude Explains, a new dedicated page on its website that's generated mostly by the company's AI model family, Claude.
techcrunch.com/2025/06/03/anthropics-ai-is-writing-its-own-blog-with-human-oversight
Real TikTokers are pretending to be Veo 3 AI creations for fun, attention
From music videos to “Are you a prompt?” stunts, “real” videos are presenting as AI.
arstechnica.com/ai/2025/05/real-tiktokers-are-pretending-to-be-veo-3-ai-creations-for-fun-attention
The OpenAI board drama is reportedly turning into a movie | TechCrunch
A film that will portray the chaotic time at OpenAI, when co-founder and CEO Sam Altman was fired and rehired within a span of just five days, is A film that will portray the chaotic time at OpenAI, when co-founder and CEO Sam Altman was both fired and rehired within a span of just five days, is reportedly in the works.
techcrunch.com/2025/06/03/the-openai-board-drama-is-reportedly-turning-into-a-movie
The Gmail app will now create AI summaries whether you want them or not
Workspace users will be seeing a lot more of Google’s AI summaries soon.
arstechnica.com/google/2025/05/the-gmail-app-will-now-create-ai-summaries-whether-you-want-them-or-not
What’s next for AI and math
The last year has seen rapid progress in the ability of large language models to tackle math at high school level and beyond. Is AI closing in on human mathematicians?
www.technologyreview.com/2025/06/04/1117753/whats-next-for-ai-and-math
DeepSeek may have used Google's Gemini to train its latest model | TechCrunch
Chinese AI lab DeepSeek released an updated version of its R1 reasoning model that performs well on a number of math and coding benchmarks. Some AI researchers speculate that at least a portion came from Google's Gemini family of AI.
techcrunch.com/2025/06/03/deepseek-may-have-used-googles-gemini-to-train-its-latest-model
Trump bans sales of chip design software to China
Move is another attempt to make it tougher for China to develop cutting-edge AI hardware.
arstechnica.com/tech-policy/2025/05/trump-bans-sales-of-chip-design-software-to-china
Inside the effort to tally AI’s energy appetite
Takeaways from our data-driven story on how and why Big Tech is aiming to reshape our energy grids around artificial intelligence.
www.technologyreview.com/2025/06/03/1117685/inside-the-tedious-effort-to-tally-ais-energy-appetite
AI video just took a startling leap in realism. Are we doomed?
Google’s Veo 3 delivers AI videos of realistic people with sound and music. We put it to the test.
arstechnica.com/ai/2025/05/ai-video-just-took-a-startling-leap-in-realism-are-we-doomed
This benchmark used Reddit’s AITA to test how much AI models suck up to us
The new benchmark, called Elephant, makes it easier to spot when AI models are being overly sycophantic—but there’s no current fix.
www.technologyreview.com/2025/05/30/1117551/this-benchmark-used-reddits-aita-to-test-how-much-ai-models-suck-up-to-us
It’s too expensive to fight every AI copyright battle, Getty CEO says
Getty dumped “millions and millions” into just one AI copyright fight, CEO says.
arstechnica.com/tech-policy/2025/05/extraordinarily-expensive-costs-force-getty-to-pick-its-ai-legal-battles
The AI Hype Index: College students are hooked on ChatGPT
MIT Technology Review’s highly subjective take on the latest buzz about AI, including Meta’s AI friends and how generative AI is helping a Neuralink patient communicate faster.
www.technologyreview.com/2025/05/28/1117468/ai-hype-index-college-students-chatgpt-meta-apple-anthropic-grok
Reply