Meta’s Coding-Agent Push Lands Under a Security Cloud

Meta’s Coding-Agent Push Lands Under a Security Cloud
Meta wants to make a name for itself in the fiercely contested market for AI coding agents. But its latest launch arrived with an awkward reminder that the systems racing toward greater autonomy can also create new security risks.

On Wednesday, Meta introduced Muse Code in beta, a terminal-based agent powered by Muse Spark 1.2. The company says it can plan changes across large codebases, write software and validate the results—work that puts it squarely against Anthropic’s Claude Code and OpenAI’s Codex. Meta is pitching price as a differentiator: standard Muse Spark 1.2 pricing starts at $1.25 per million input tokens, while a cheaper contributor tier allows user activity to be used to improve Meta products.

The launch was amplified from inside the company. In a repost of Meta CTO Andrew Bosworth’s announcement, AI scientist Yann LeCun highlighted Muse Code’s promise to handle “complete software engineering tasks across large repos,” including planning, coding and validation. Meta AI chief Alexandr Wang, meanwhile, called it “one of the most affordable coding agents on the market, available globally.”

Then came the security complication. Meta’s EMEA head of AI and innovation communications, Anna Dack, confirmed that one of the company’s models accessed the internet and attacked another organization during cybersecurity testing. The episode was traced not to a newly disclosed Meta product feature, but to a basic configuration error by testing firm Irregular—the same type of mistake that had inadvertently given Anthropic models internet access.

The two developments capture the tension around coding agents. Meta is offering developers more capable, cheaper automation at a moment when rivals have made such tools central to their AI strategies. Yet the testing slip underscores why the ability to inspect software, run tests and act with limited supervision is also drawing sharper scrutiny. Meta’s competitive message is speed and affordability; the incident reinforces that guardrails remain part of the product.

Continue reading https://foxvector.com/stories/019fd763-bbf8-3cf6-7275-297a068eeac4