Meta updates AI safety with Llama Guard 4 and LlamaFirewall
Meta is launching new safety tools for Llama, including the multimodal Llama Guard 4, LlamaFirewall, and more efficient Prompt Guard 2 models via GitHub and Hugging Face.

What happened?
Meta has launched a suite of new and updated safety tools for the Llama ecosystem, including Llama Guard 4, which supports both text and image processing. The company is also introducing LlamaFirewall as a central security hub against prompt injection and insecure code generation, alongside two new versions of Prompt Guard 2 (86M and 22M) to detect jailbreak attempts. The models and tools are available via GitHub, Hugging Face, and Meta's safety portal.
Key facts
| Lanseringsdatum | 9 augusti 2025 |
|---|---|
| Resursminskning Prompt Guard 2 22M | Upp till 75 % lägre latens och beräkningskostnad |
| Nya Prompt Guard-storlekar | 86M och 22M parametrar |
Why it matters
As AI applications become increasingly multimodal and handle more external integrations, the risk of security threats such as manipulated instructions and malicious plug-ins grows. The smaller Prompt Guard 2 22M model also reduces latency and computational costs by up to 75 percent, making advanced protective filtering more accessible for real-time environments.
Who is affected?
The tools are primarily aimed at AI developers and cybersecurity teams building applications based on Llama models. This affects both smaller developers with limited computational budgets and larger enterprises requiring advanced multimodal protection.
Impact on the EU
The tools are distributed globally via open platforms such as GitHub and Hugging Face, making them immediately available to developers within the EU. However, the associated Llama API service, which includes Llama Guard 4, is currently limited to a preview.
What else you should know
CyberSec Eval 4 has also been updated to assist organizations in measuring how well AI systems perform in cybersecurity-related tasks. The source material does not specify exact dates for a broader rollout of the Llama API.
Quick answers about this story
Vad har hänt?
När hände det?
Varför spelar det roll?
Är verktygen tillgängliga i Europa?
The link opens in a new window and leads to the publisher's own site.
Källan har spårats automatiskt från utgivaren via Aheadlines signalkedja.
AI-verktyg i artikeln
Topics
Get similar news straight to your inbox
The reader's room
Send in a question or an addition. The newsroom reads everything before it's published and replies when relevant. No AI-generated text – just people.
Sign in to submit a comment or question.
Read the article through your role
- Decide whether this affects strategy over 6–12 months or is just noise.
- Discuss with leadership: do we own the right question or does ownership need to move?
- Ask: what risk are we taking by NOT acting on this this quarter?
Generated angle — not editorial analysis of "Meta updates AI safety with Llama Guard 4 and LlamaFirewall"