Meta announced the launch of Muse Spark 1.1, a new artificial intelligence model developed by the Meta Superintelligence Lab. Although it belongs to the same product line presented in April, version 1.1 represents a significant advancement, specifically aimed at AI agents, with an emphasis on programming, logic, and task automation.
The company aims to deliver more complex and comprehensive functionalities with minimal need for human intervention. This model is initially available as a preview for developers located in the United States, seeking to reduce the technological gap compared to its competitors. Although there are no independent tests yet to validate its performance, benchmarks released by Meta itself indicate that Muse Spark 1.1 surpasses GPT-5.5, Gemini 3.1 Pro, and Opus 4.8 across various metrics.
The main differentiator of Muse Spark 1.1 lies in its enhanced ability to operate as an autonomous agent. The update focuses efforts on three main areas: programming, computer usage automation, and multimodal reasoning. The latter allows the system to simultaneously process and correlate data from texts, images, videos, and files.
To support these functions, the system has a context window reaching up to 1 million tokens. This enables Spark to maintain memory of actions during extensive work sessions or retrieve information from early stages of large projects.
In the field of software development, the tool demonstrates notable improvements in bug detection and correction. In internal tests, Muse Spark 1.1 was capable of creating a chat application from scratch, diagnosing interface issues by capturing screen images, and even modifying the original source code to implement fixes, all without any human assistance.
Beyond the coding environment, the AI can also navigate the internet autonomously and control a device's operating system. A practical example of this was when the assistant extracted relevant information from a product video and drafted a complete advertisement for Facebook Marketplace using the user's browser.
According to information obtained by the Reuters agency, access to Muse Spark 1.1 has a lower cost than leading market models, such as Claude Sonnet 4.6. A comparison of costs per million tokens shows that Muse Spark 1.1 costs US$ 1.25 (R$ 6.42) for input and US$ 4.25 (R$ 21.83) for output, while Claude Sonnet 4.6 costs US$ 3 (R$ 15.41) for input and US$ 15 (R$ 77.04) for output.
Mark Zuckerberg, the company's CEO, shared the news on X/Twitter, stating that the company's current focus is on offering robust models with independent and multimodal agents, but at a very accessible price.
The new model is already accessible in the Thinking mode of Meta AI and is scheduled for integration soon into WhatsApp, Instagram, and Facebook chatbots, as well as the company's own smart glasses. To mitigate security concerns, Meta emphasized that the model underwent rigorous evaluations, demonstrating high resilience against code injection attacks and hallucination occurrences.
It is worth noting that this announcement occurred in the same week that Meta presented Muse Image, whose first image generation model sparked discussions due to the use of Instagram user posts as a basis for its creations.
Meta has introduced a new artificial intelligence-based tool capable of generating images using public photos from Instagram users. To disable this feature, users must find a hidden toggle switch deep within the settings menu.
The tool, named Muse Image, became available on Tuesday through the Meta AI app, meta.ai, and in Instagram Stories in the US, and in some countries, also in WhatsApp. It allows users to tag a friend's account on Instagram to include their public pictures in AI-generated images. Meta positions this tool as 'AI serving social experiences that billions of people already love.'
A particular concern arises with Instagram users from South Africa and other individuals outside the country: all public accounts are automatically activated. Private profiles and users under 18 are excluded by default, but any owner of a public profile provides their photos for AI remixing unless they manually turn off this option.
For South Africa, where Instagram is a primary platform for photographers, small businesses, influencers, and private individuals who post pictures, this represents a very broad scope of application. Disabling this feature is located in Settings, then Sharing and Remixing, under the section 'Allow people to use your content in Instagram and with Meta AI features,' where toggles for Posts and Reels must be switched off. Even after disabling the feature, images previously created using the user's photos remain active. Users do not receive notifications about who uses their content, nor is there a way to see what has already been generated.
Talent agency CAA, which represents stars such as Tom Cruise, Brad Pitt, Zendaya, and Meryl Streep, issued a statement on Wednesday evening, urging Meta to change the default settings. CAA stated: 'No one's name, image, likeness, or creative work should be used by any third party, including AI models, without clear, documented consent.' The agency demanded that protection be the default setting in Muse Image, not an exception, and that people could voluntarily agree to have their image used for AI content creation.
The CAA statement also called for clear information, prompt removal of unauthorized content, and the ability for creators to 'impose restrictions, control usage, and prevent unauthorized approval or exploitation.' Meta did not respond to requests for comment.
This is not the first instance where a tech giant places the burden of obtaining consent on the user. OpenAI took a similar step with Sora 2, its AI video platform, which allowed the creation of videos featuring celebrities and movie characters without permission. CAA and the Motion Picture Association condemned this move. OpenAI CEO Sam Altman promised 'more detailed control' over character generation. However, three months later, OpenAI completely shut down Sora and terminated a $1 billion deal (18.3 billion rounds) with Disney.
The Silicon Valley concept regarding permission requests still implies that you have consented unless you find a switch. Whether Meta will follow this same scenario depends on how much public outcry it generates. Currently, Instagram users in South Africa with public accounts are at risk regardless of their wishes. If they want to exit this situation, the path lies through Settings, Sharing and Remixing, and two toggles. No one is telling them where to find them.
Meta has introduced Muse Image, a new artificial intelligence model integrated into Instagram. This feature allows Meta AI to use photos from public profiles as a basis for generating new images.
According to Wired, Muse Image, developed by Meta Superintelligence Labs, is part of Meta's strategy to compete in the AI image creation market. The mechanism is simple: by referencing a public profile in a creation prompt, any individual can request an image inspired by that user's appearance.
The company claims that the tool's purpose is to expand creative capabilities, citing examples such as producing personalized invitations, developing collaborative visual concepts, and generating graphics from publicly accessible images. Meta itself stated that tagging a username allows Meta AI to use public photos to create an image ready for dissemination, whether for an invitation or a graphic.
This feature is automatically activated for public accounts. Therefore, those who wish to prevent such use must look for the disabling option in the account settings. The types of content that can serve as references include photos posted on open profiles, images shared on Instagram, videos published on Reels, and other content used in Meta AI creations.
This permission modification can be done directly in the Instagram app without needing to make the account private. The user must access their profile, tap the three lines in the upper right corner, and navigate to the 'Sharing and reuse' section. There, they need to find and adjust the option titled 'Allow people to use your content on Instagram and with Meta AI features,' where they can disable the use of posts and Reels in future tool creations.
Instagram's Help Center clarifies that if the default setting is maintained, 'people may create content using your Instagram content through Meta AI features.' It is important to note that disabling this option only prevents future use of the content; it does not delete the AI images that were previously generated using the user's photos.
Additionally, Instagram does not notify users when third parties use Meta's features to create images from another person's public content. The platform's help page confirms: 'You will not be notified about content created with Meta AI features.' The launch of Muse Image highlights the integration of generative tools into social networks, raising user concerns about control over their own images.
