--- title: OpenAI folds safety into research, Meta pulls Muse Image... description: OpenAI merged its safety team into research under a new VP role, continuing a pattern of structural safety downgrades that critics say reduces its... canonical: https://daily.dev/posts/openai-folds-safety-into-research-meta-pulls-muse-image-ai-after-backlash-lri7h4o9o twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: OpenAI folds safety into research, Meta pulls Muse Image AI after backlash | daily.dev og:description: OpenAI merged its safety team into research under a new VP role, continuing a pattern of structural safety downgrades that critics say reduces its... og:url: https://daily.dev/posts/openai-folds-safety-into-research-meta-pulls-muse-image-ai-after-backlash-lri7h4o9o og:image: https://api.daily.dev/og/posts/lRI7h4O9o.png og:image:alt: OpenAI folds safety into research, Meta pulls Muse Image AI after backlash og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # OpenAI folds safety into research, Meta pulls Muse Image AI after backlash **[Agentic Digest](https://daily.dev/sources/agents_digest)** · 4 min read · 2 upvotes · 0 comments ## Summary OpenAI merged its safety team into research under a new VP role, continuing a pattern of structural safety downgrades that critics say reduces its independence. Meta launched and then yanked Muse Image AI within three days after SAG-AFTRA and talent agencies objected to opt-out-by-default likeness generation. The AI price war is sharpening: GPT-5.6, Grok 4.5, and Muse Spark 1.1 all launched with cost as the headline, while Anthropic is moving the opposite direction. A new Ghostcommit attack hides prompt injection in PNG images to steal secrets from AI code reviewers. ## Content **TLDR:** OpenAI merged its safety team into research under a new VP role, continuing a pattern of structural safety downgrades that critics say reduces its independence. Meta launched and then yanked Muse Image AI within three days after SAG-AFTRA and talent agencies objected to opt-out-by-default likeness generation. The AI price war is sharpening: GPT-5.6, Grok 4.5, and Muse Spark 1.1 all launched with cost as the headline, while Anthropic is moving the opposite direction. A new Ghostcommit attack hides prompt injection in PNG images to steal secrets from AI code reviewers. --- ## OpenAI folds safety team into research for the second time in two years OpenAI is merging its safety and research organizations under a new VP of Research and Safety role held by Mia Glaese, following the departure of head of safety systems Johannes Heidecke. This is the second restructuring of this kind in under two years — the Superalignment team dissolved in 2024, the Mission Alignment team in early 2026. Critics argue the move reduces safety's structural ability to delay or block product launches. The changes land while 42 state attorneys general are investigating OpenAI and the company has filed a confidential IPO. [Read more](https://daily.dev/posts/IBmMrNMIG) ## Meta pulls Muse Image AI three days after launch over likeness consent backlash Meta's Muse Image AI, the first product from Meta Superintelligence Labs, let users generate images referencing any public Instagram account by default — no opt-in required. SAG-AFTRA, talent agency CAA, and actors including Hannah Einbinder publicly objected. Meta initially defended the feature, then reversed within 24 hours of Mark Zuckerberg's public pushback, acknowledging it 'missed the mark' on privacy. The episode fits a recurring pattern: Meta ships AI features with opt-out defaults and retreats under pressure. [Read more](https://daily.dev/posts/NJ4Q4rd3Y) ## Ghostcommit hides prompt injection in PNG images to steal secrets from AI code reviewers Researchers from the ASSET Research Group demonstrated an attack that embeds prompt injection instructions inside a PNG image referenced by an AGENTS.md file. AI code reviewers like CodeRabbit and Bugbot never open image files, so the malicious PR passes review. When a developer later asks a coding agent for a routine task, the agent reads AGENTS.md, follows the pointer to the image, opens the repo's .env file, and encodes every secret as a list of integers in generated code — bypassing secret scanners that don't decode integer tuples back to ASCII. Cursor and Antigravity leaked secrets under multiple models; Claude Code refused every time. [Read more](https://daily.dev/posts/tpKlXUCTC) ## AI price war sharpens as GPT-5.6, Grok 4.5, and Muse Spark 1.1 compete on cost Within a single week, OpenAI, xAI, and Meta all launched frontier models with pricing as the primary pitch. Muse Spark 1.1 is available at $1.25/million input tokens and $4.25/million output tokens; a cost comparison circulating today puts an $8,545 Anthropic Opus bill at roughly $1,000 using GLM 5.2 via OpenRouter. Anthropic is taking the opposite bet, raising prices on its top coding model. For teams running agentic workloads at 150K–200K tokens per task, routing by cost-per-finished-task rather than cost-per-token is becoming the real optimization target. [Read more](https://daily.dev/feed-by-ids?id=efcOThYLo&id=rzsy72oV4) --- ## Also notable - **Muse Spark 1.1 scores 48/70 across 7 benchmark tasks, tops CritPT:** Independent testing placed Muse Spark 1.1 at 68.57% (48/70) across UI generation, Three.js, SVG, math, and agentic workflows — 6th on the leaderboard overall — with standout performance on hard math and complex agentic tasks, but inconsistent logic on UI tasks and poor file-awareness in parallel agent setups. [Read more](https://daily.dev/feed-by-ids?id=SNam31Pp7&id=GYC4V4jrL&id=qA9dvtjJn) - **Veracode: 45% of AI-generated code samples contain security vulnerabilities, Java hits 72% failure rate:** Veracode's 2025 GenAI Code Security Report found nearly half of AI-generated code samples are vulnerable, with Java the worst offender at 72%, while a Stanford study found developers using AI coding assistants wrote less secure code while feeling more confident about it. [Read more](https://daily.dev/posts/UszaxywBI) - **Reame CPU-first LLM inference server benchmarks at 4.8x end-to-end speedup with warm disk cache:** Reame, built on llama.cpp for cheap hardware like free-tier VMs, achieves 4.8x end-to-end speedup with a warm disk KV cache, 3.2x with speculative decoding, and 2.3x with archive speculation — explicitly positioning against Ollama by getting faster the longer it runs. [Read more](https://daily.dev/posts/W9j1C37BO) - **Tencent in talks to become Manus' largest shareholder after China forced Meta to unwind its $2B acquisition:** Beijing used its foreign-investment security review mechanism to block Meta's ownership of Manus despite the company having relocated to Singapore, with Tencent now reportedly in negotiations to take the largest shareholder position. [Read more](https://daily.dev/posts/ZIwLll341) - **OpenAI's first hardware device, a camera-equipped smart speaker, won't ship before February 2027:** The Information reports the device can ship no earlier than February 2027, while Apple is separately seeking legal restrictions covering disputed designs, manufacturing methods, and suppliers tied to the ongoing trade secrets case. [Read more](https://daily.dev/posts/IJcyNGiQP) ## Similar posts on daily.dev - [Meta pulls Muse Image AI after Hollywood privacy backlash](https://daily.dev/posts/meta-pulls-muse-image-ai-after-hollywood-privacy-backlash-nj4q4rd3y) · The Next Web · 2 upvotes · 1 comments - [Anthropic’s answer to Dots and Muse is already inside Claude](https://daily.dev/posts/anthropic-s-answer-to-dots-and-muse-is-already-inside-claude-i05o6a1ds) · The New Stack · 0 upvotes · 0 comments - [Meta enters the crowded AI coding battle with Muse Spark 1.1](https://daily.dev/posts/meta-enters-the-crowded-ai-coding-battle-with-muse-spark-1-1-wtfjrleyt) · TechCrunch · 1 upvotes · 0 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#openai](https://daily.dev/tags/openai), [#ai-safety](https://daily.dev/tags/ai-safety), [#prompt-injection](https://daily.dev/tags/prompt-injection) [View this post on daily.dev](https://daily.dev/posts/openai-folds-safety-into-research-meta-pulls-muse-image-ai-after-backlash-lri7h4o9o) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"DiscussionForumPosting","mainEntityOfPage":"https://daily.dev/posts/openai-folds-safety-into-research-meta-pulls-muse-image-ai-after-backlash-lri7h4o9o","headline":"OpenAI folds safety into research, Meta pulls Muse Image AI after backlash","text":"OpenAI merged its safety team into research under a new VP role, continuing a pattern of structural safety downgrades that critics say reduces its independence. Meta launched and then yanked Muse Image AI within three days after SAG-AFTRA and talent agencies objected to opt-out-by-default likeness generation. The AI price war is sharpening: GPT-5.6, Grok 4.5, and Muse Spark 1.1 all launched with cost as the headline, while Anthropic is moving the opposite direction. A new Ghostcommit attack hides prompt injection in PNG images to steal secrets from AI code reviewers.","url":"https://daily.dev/posts/openai-folds-safety-into-research-meta-pulls-muse-image-ai-after-backlash-lri7h4o9o","datePublished":"2026-07-12T04:18:24.608Z","dateModified":"2026-09-13T21:14:04.453Z","author":{"@type":"Organization","name":"Agentic Digest","logo":"https://media.daily.dev/image/upload/s--V91DY4ls--/f_auto,q_auto/v1772617267/logos/agents_digest","url":"https://daily.dev/sources/agents_digest"},"interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":2},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"isPartOf":{"@type":"WebPage","url":"https://daily.dev/sources/agents_digest","name":"Agentic Digest"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Agentic Digest","item":"https://daily.dev/sources/agents_digest"},{"@type":"ListItem","position":3,"name":"OpenAI folds safety into research, Meta pulls Muse Image AI after backlash"}]} ```