In a reversal, ChatGPT now refuses to do many things that Claude is happy to address. (Examples from both GPT-5 Pro and GPT-5 Thinking)
SAFETY
-
Science integrity compromised by opportunists and ideologues
By
–
Science requires the adherence to a special code, a commitment to seek and serve truth above all other things. By opening up science to opportunists, impostors and ideologues, we defiled it, robbed it of its status and sanctity, and deprived ourselves of its fruits
-
AI Story Generation: Exploring What Machines Writing Means
By
–
This is an interesting debate about AI text between an OpenAI researcher who thinks about AI writing and one of the great short story masters Now that we have machines that can write stories, occasionally very good or moving stories, we need to think more about what that means
-
Sora 2 integrated into Photo AI with audio support
By
–
✨ Okay @OpenAI Sora 2 is live inside Photo AI!!!😊
— @levelsio (@levelsio) 6 octobre 2025
Kinda because I can't get image input with people to work yet, the safety filter blocks most, but hopefully that's fixed soon
But it follows the prompt at least, and has audio, so it's a start
And no watermarks now!!! https://t.co/E83lEJ2BSA pic.twitter.com/ivx5OlMTGTOkay @OpenAI Sora 2 is live inside Photo AI!!! Kinda because I can't get image input with people to work yet, the safety filter blocks most, but hopefully that's fixed soon But it follows the prompt at least, and has audio, so it's a start And no watermarks now!!!
-
Image blocking issues with AI content moderation systems
By
–
Trying!!! But everything is getting blocked, tried both pics, even a wheat field gets blocked! :O
-
Petri Framework Advances AI Model Safety Assessments
By
–
Petri builds on our alignment assessments in the Claude 4 and 4.5 System Cards; the @AISecurityInst also successfully built on a pre-release version of Petri for their assessments of our models.
-
Petri Open-Source AI Safety Research Tool Released
By
–
Petri is open-source and available now: http://
github.com/safety-researc
h/petri
… Read the full technical report: https://
alignment.anthropic.com/2025/petri -
Anthropic Open-Sources AI Audit Tool for Claude Sonnet 4.5
By
–
Last week we released Claude Sonnet 4.5. As part of our alignment testing, we used a new tool to run automated audits for behaviors like sycophancy and deception. Now we’re open-sourcing the tool to run those audits.
-
Petri: Automated Agent Tool Audits AI Models Across Scenarios
By
–
It’s called Petri: Parallel Exploration Tool for Risky Interactions. It uses automated agents to audit models across diverse scenarios. Describe a scenario, and Petri handles the environment simulation, conversations, and analyses in minutes. Read more:
-
CiteCheck Benchmark Standards for Precise Claim Verification
By
–
The goal of a CiteCheck benchmark should not be to check, "Does the cited document say something *related* to the claim?", but "Does the document state or very directly support the *exact* claim it's being cited about?" Here are two failures from the last generation of LLMs that