CRA compliance hardens code and moves companies closer to digital sovereignty: the CRA mandates controls manufacturers must follow (non-commercial open-source maintainers are exempt), OpenSSF's Baseline translates regulations like the CRA into tiered practices, and open source lets organizations inspect, fork and own their stack rather than depend on a single vendor.
Akrites is a shared security incident response team for critical-infrastructure stakeholders. It provides a confidential, structured setting for coordinating vulnerability discovery, deduplication, validation, remediation, and synchronized disclosure across the open-source projects on which those stakeholders depend.
Generally AI is an InfoQ podcast about artificial intelligence, machine learning, and data engineering. Hosted by Anthony Alford and Roland Meertens, it presents conversations and discussions on topics including large language models, generative AI, algorithms, programming, and AI's relationship with the physical world. InfoQ maintains an episode archive and RSS feed for the podcast.
InfoQ Online Certification Programs is a set of five-week professional-development cohorts for senior software engineers, architects, and technical leaders, covering software architecture, AI engineering, AI security and privacy, AI-assisted engineering, and high-performing teams. InfoQ adds and schedules cohorts on its online program site. Each week, participants watch a QCon talk, join a four-hour live session with a facilitator and a confidential peer group, and apply a framework from the talk to a decision from their own work. The program includes time-boxed homework, a capstone technical article presented in the fifth week, and the possibility of publication on InfoQ. Certification requires consistent attendance and active participation in the live sessions, along with successful completion of the group capstone project; sessions are not recorded to preserve confidentiality.
InfoQ Podcasts is a maintained collection of weekly podcasts featuring conversations and discussions with software architects, senior developers, technical team leaders, engineering-management practitioners, and other senior software leaders. Episodes cover software architecture, AI and data engineering, software development, engineering practices, DevOps, cybersecurity, engineering culture, and management, with podcast subscriptions and interview transcripts available through InfoQ.
The OpenSSF Cyber Reasoning System is a harness being developed by the Open Source Security Foundation for AI-assisted software development. It is designed to let users select models, configure how those models are used, and provide skills and additional context to AI coding agents. Planned skills include OpenSSF Scorecard, SLSA, and Security Insights, directing agents toward established software security practices.
The OpenSSF Open Source Software Cyber Reasoning System is a harness for AI-assisted software security. It plugs in selected AI models, tunes how they are used, and applies skills and contextual information to guide code development toward established security practices.
OpenSSF Scorecard is an automated open-source software security assessment tool developed by the Open Source Security Foundation (OpenSSF). It evaluates projects and dependencies for software supply-chain risks using checks covering branch protection, code review, pinned dependencies, signed releases, token permissions, vulnerability status, fuzzing, security policies, build processes, testing, and project maintenance. Each check receives a score from 0 to 10, with results combined into an aggregate security-posture or risk-weighted score alongside individual findings and remediation guidance. Scorecard is available as a command-line tool, Docker container, GitHub Action, REST API, web viewer, and badge generator; it can assess repositories and package sources from GitHub and GitLab, and maintainers can run it on repository changes and publish results.
QCon San Francisco 2026 is an in-person software engineering conference and training event for senior engineers, architects, and technical leaders, held November 16–20, 2026, in San Francisco. The conference covers applied AI and machine learning, agent orchestration, evaluation and guardrails, API design, architecture, distributed systems, platform engineering, observability and resilience, developer experience, and engineering organizations. Its program is organized into 12 curated tracks with more than 60 presentations from senior practitioners describing systems and practices used in production. Optional half- and full-day training sessions run November 19–20 and cover topics including building AI agents and selecting among agent-development SDKs.
The Software Architects' Newsletter is a monthly InfoQ newsletter covering the frameworks, platforms, tools, technologies, techniques, and architectural practices relevant to software architects and aspiring architects. It is curated by InfoQ and domain experts, and provides actionable takeaways, insights from software practitioners, and lessons about current architecture work. Issues are delivered once a month and past issues are available on the newsletter page.
Searchable transcript of Getting CRA Compliant Hardens the Code and Gets Your Company Closer to Digital Sovereignty — InfoQ (27:04). Search for a phrase, then click its timestamp to jump straight to that moment in the video.
Captions sourced from the original video on YouTube, published by InfoQ. The video, its captions and all related intellectual property remain the property of their respective owners; AINotes claims no ownership. Provided for research, accessibility and search — see the Transcript Notice and Copyright Policy.
00:00 How much authority should an AI agent have? At CUCON AI New York, learn how senior practitioners set permissions and evaluate agents in production. Use their experience to decide whether your own agents need boundaries and checks before release. Join us December 15th and 16th, 2026 at the Western Jersey City Newport. Explore the program at new york.con.ai.
00:29 AI. >> Hello everybody. I'm Olymp editor and I have in front of me Christopher Robinson or how his handle goes these days is crop and we'll let him introduce himself and provide more details about why. >> Hello everybody. It's thank you for having me today. My name is Chris Robinson also known as Crobe. I am the CTO and chief security architect for a couple Linux foundation projects.
01:02 Firstly, the open source security foundation which focuses in on application security, supply chain security and AI security predominantly trying to help support upstream maintainers. Then my other role is I get to help lead a new project called Credies which is an upstream open-source coordinated incident response team. >> Okay, that sounds like a lot.
01:22 You have multiple hats as as people see in presentations. Yes. [laughter] Yeah. But open SSF is definitely a stronghold for security and >> it's BA probably B AI. Uh so before AI things happened, [laughter] >> people were pushing a lot for having security moving to the left meaning >> yes >> having all those closer to the development team because initially what happened is that security was left behind with a lot of other stuff that are quite important documentation and so on so forth.
02:02 So at least in my experience whenever something bad was happening you had your security guy your security architect scrambling down and trying to do something and then trying to convince the developers to do it. Now as we are recording this at the beginning of November and we saw in the last one two weeks a lot of people scrambling around due to PHP and probably others as well because incidents are rising and just looking at the issues that the guys from PHP had like well we never thought that all these four conditions
02:34 will happen at some point well they did AI is helping with that. >> Yes helping. Yeah, exactly. Exactly. When and who should come to open SSF? What's your targeted audience? So, we have two primary personas that we try to interact with. Firstly is the upstream open-source developer maintainer. So we have a ton of guidance, best practices, tooling and training to help teach software engineers like what the expected security controls and practices and things you need to do to harden your project and your infrastructure.
03:11 So that's kind of our our primary, you know, customer, so to speak. And then secondarily, we're also really concerned about everyone that's consuming those projects. Sometimes it might be another open source maintainer that's integrating another project as a dependency but predominantly it's a lot of downstream manufacturers and enterprises like banks and hospitals, retailers, you know, whatever it might be just running their business.
03:37 So we have also an additional set of tools and guidance to help those consumers evaluate and understand the security qualities of the project that they're trying to potentially integrate into their enterprise which you know we had talked right before we started recording about your use of scorecard and that's one of the tools that's kind of a dual purpose.
04:00 It can help upstream maintainers, you know, choose the right dependencies or if they're interested in kind of showing off their security expertise and bragging a little bit, scorecards away. It's a very strong public signal that they can put out there showing that I follow good practices within the security space. And then consumers can look at for those signals and make decisions.
04:21 you know, if there's two packages that they are considering to integrate into a solution, they might want to pick one that might have stronger security posture potentially. >> Yeah. And that's why I mentioned even before we started recording about the scorecards is that they're not looking only at the technical aspect, right? >> Because we we saw that a lot of the things that happened were on the social engineering part or not being careful.
04:46 So I'm just thinking now about typos squatting or doing a bit of leveraging on especially in the in the Python ecosystem where they just went with the highest possible version and then that just exploded everything and that's quite important these days especially I think people can say that it's something that happened in the last weeks but my feeling is that it happened ever since the pandemic hit.
05:12 During the pandemic a lot of people were just bugging in and they were just working on projects and after that I have in mind a couple of the reports published by different companies that are working in the supply chain space for instance sonata type I followed closely and you saw a lot of decrease into the maintenance of open source project and then obviously those opened a wide door for attack vectors.
05:41 >> Yep. Other than scorecards, what will be a tool that you'll see out and that's also kudos for for open SSF is because a lot of people are coming with best practices and we all know what's good or we know where to find the information but the problem is how do you transform that? That's the the important part that it's hard to to put together. So what other tool would you just single out?
06:08 I'm going to cheat and I'm going to say two. >> Excellent. Great. >> Firstly, I would say our security baseline which is kind of the next evolution of what scorecard is. So baseline has 41 recommended security things to do. In cyber security we call them controls or practices. What we've done is we've looked across the globe at international laws like the EUR.
06:39 We've looked at things like the uh NIST secure software development framework. So we've looked at all these different laws and frameworks and we put together this reasonable list of things a project could do or that a consumer should look for like you mentioned. It's a combination of technical things like you should have this setting on your source code repository to prevent you know branch protection for example.
07:03 And there's also more softer controls like do you have a published security policy? Now that's just it's just a written document that describes what the project does. Do you have a documentation on how you onboard new contributors and how do you move up the maintainer ladder? So baseline's broken up into three levels of maturity. The first level which is the simplest which would be kind of analogous to kind of a mid-range scorecard score.
07:32 We feel that would take a experienced software developer two hours or so to implement this stuff. And the first level is predominantly documentation. And what that allows you to do is again it kind of coaches you where upstream maintainers generally aren't required to follow the cyber security laws and frameworks. Like the EU CRA is very explicit that an open source maintainer if you're not making a profit on your work you are not beholden to follow the requirements of the law but manufacturers are and you know
08:12 governments care a lot about that. So what the baseline does is it has this kind of legal regulatory flavor to it that I, you know, I don't expect an upstream maintainer to go read the CRA or to go read US executive orders or go read legislation in Japan. But the baseline does that for you and kind of says based off of you know how modern software should be developed in the expected security practices here's a tiered way that you could implement this which ideally makes your project more interesting for downstream
08:43 which you know ideally draws in more contributors it draws in more users which helps give you more feedback for RFS maybe gives you more eyes looking at your codebase to help improve it now the other tool I'll suggest is a framework called Salsa, which is a supply chain, a pedigree and providence, an integrity specification. And so you combine that with something like a scorecard score or you're at a baseline level.
09:14 And Salsa is thinking about all of the infrastructure that is used by the project or the developer. So your IDE environment, your workstation, it thinks about your source code repository system, your CI/CD system, your uh distribution system and provides again different levels of maturity and ideally the higher up you are in that maturity curve, the more hardened that infrastructure is.
09:41 And therefore you can take something like a scorecard or baseline that's kind of focused around the process and the code and then salsa kind of overlays on top of that and tells you how all that software and code and practices integrate through and move through the supply chain and how relatively secure you should expect that infrastructure to be. >> You you did mention the AI world among the interest you have these days.
10:08 So I have to ask the question uh but it's all nice and we would expect or yesterday would have we would have expected to have um human audience but now day we are just priding oursel that the skills that we managed to gather throughout decades we just forgotten in I don't know a couple of months and now we are just all prompters we are all um forward engineers or backwards engineers as you ought to put it >> how is it is to include all this tooling into harnesses because now all our experience transformed itself into
10:49 harnesses. It's really interesting. I I get the chance to talk a lot about AI security and broadly LLMs and agents and all this new AI technology has really pulled in a ton of new stakeholders into the software engineering software development practice where you know somebody like me that has very minimal software engineering training I could write a program now with very simple prompting and queuing and ideally I've done things like you're hinting at.
11:21 I have skills and I have the harness that's trained and able to leverage these best practices. But that's the challenge we have is we not only have the the regular kind of experienced software engineers that have been participating, but now we have this whole slew of people from accounting and marketing and just literally everywhere within a business able to write software.
11:45 And where things have fallen down is that AI and especially with these new um Frontier LMS, it's like your birthday or it's like Christmas. Everyone's so very excited to open the package and play with the new toy. Everybody's racing forward to play with these new amazing tools and kind of think about how they can integrate those capabilities, but they've, as you mentioned, kind of forgotten.
12:10 We have 50 years of software engineering discipline and practice that have helped get us for better or worse where we are today and AI is kind of allowing people to ignore some of those things and that's where the open SSF has has been focused this year. We brought in a new security architect for AI and she's helping us deploy a new system called the open- source software cyber reasoning system which is exactly a harness.
12:38 It allows you to plug in whatever models you want. It allows you to tune how those are leveraged and allows you to apply skills and additional context. And that's where my foundation is spending. We've spent the back half of 2026 focusing on this. And then 2027, my goal is to have skills and other tools to plug into harnesses. So next year you'll see a scorecard skill and you'll see a salsa skill and you'll see a security insight skill.
13:09 So we'll have all these things that enable the robots to move along their merry way but ideally challenge them and force them to use these known proven security practices so the output of the code is ultimately ideally more hardened and secure. >> Okay, happy to hear that. The other day I was just playing around with the repository and what I had to do is break down the recommendations what normal the scorecard will do for you.
13:41 Just break them down into pieces and then take them one by one and add them there. Well, I finally got to that. But as you mentioned, I knew I had the experience of how to use it. So then it was easy for me to integrate it. Right. >> But now it's quite easy to if you have some something like okay as you mentioned we are just now moving to the very thin layer of product vision and those people don't know what's to be expected and then it's it's easier for them if something just triggers that point.
14:11 >> But then thinking about like these LLMs in the hands of a senior experienced engineer they really can be a gamecher improving your productivity and helping remind you about the guard rails. But you have to know to plug in the skill or to do the thing. But yeah, it's I'm cautiously excited about it. >> But it's, you know, it's not without its challenges.
14:34 >> Well, it's accelerated. I mean, for me, what's useful, where I see a lot of enthusiasm is on the upper side of the spectrum. though quite experienced engineers and on the lower side of the spectrum where as you mentioned people from different departments are not reliant on the uh how should I put it mildly the arrogance a lot of the engineers have and then just okay it depends and all those hiding behind the finger when asked about the solution and on one side is the senior level technical people didn't have the
15:08 chance to code anymore and now They are not reliant on teams that are not on par with their expectation and on the other side it's about the productivity gains that happens but you did mention another uh important acronym these days at least in the side of the world where I am and that's CRA. >> Yes, I know you like your quotes and I like mine. So Europe is not good at implementing but we are good at regulating.
15:36 So CRA is one proof of of that. The DI act is another GDPR is another one. Well, you can take it or leave it, but I think it opened the door on all multiple things. >> Yeah. >> On the other point also the panel that you had and I attended not long ago was talking about the we call it the sovereignity but we can call it fragmentation also. And in the open source ecosystem we are talking about the global world, a global village, >> right?
16:06 If we take a step upwards because you're obviously looking at the whole global ecosystem, CRA is getting into force slowly starting from this autumn. It has multiple stages. Then the US used to have at some point um some executive order that started everything in terms of hardening software and then from that point on we discussed about memory safe programming languages.
16:32 I know that Australia invests a lot of money as you mentioned you have an agenda for this year and for next year maybe you can give us a summary where is the world from the perspective of the cyber security >> and I'll say I've been a practitioner for many decades and I started in banking and banking is a heavily regulated business no matter where the shop wherever the branch is it could be in the US it could be in Europe it could be in Africa doesn't matter it's very heavily regul ated and you know that helped give me
17:05 the experience and understanding of why following some rules are important and that's why like when the Europeans were working on the CRA I was really excited to see it and I understand it's not without its challenges regulation isn't fun but the CRA is advocating for things that make sense and have been advocated for for a very long time you know GDPR was very influential and what I enjoy about the European perspective is that it's so very citizen consumer focused.
17:36 You know GDPR was about protecting citizens data. The CRA is about protecting consumers from potentially harmful products and vulnerabilities in those products which is that you know that intention is wonderful and globally the world has recognized kind of the importance of the CRA and we're seeing different regions watching what's happening as we're starting to unfold as we move towards December 11th next year when the law actually goes into full force.
18:07 So I'm aware of motions in the UK, in India, Japan, Australia. Everybody's observing kind of what's going on in Europe and they're thinking about how do they do their own implementation. So I think gazing into my crystal ball, I think next year we will probably start to see some proposed legislation in other geographies that will be aligned with the CRA, at least spiritually.
18:33 And my hope is as a practitioner that is thinking about doing something once and having it implemented in multiple locations. I I hate rework. My dream is that these other nations that are looking at the CRA to implement their own protections. I hope that they follow the the intention and the execution and we don't get a lot of disjointed guidance. It'd be the worst thing in the world if Europe said you need to do a software bill of materials this way which today follows generally with how the industry and ecosystem
19:07 has been doing software bills of materials but if India for example says well we're going to do our own software bill of materials and there are 86 criteria and it's just that makes it very hard for manufacturers to be able to produce a product efficiently and then that puts more burden on the developers because that puts conflicting requirements ments potentially in front of them that they need to deliver on.
19:30 So my hope is next year we'll start to see other people start to move towards a CRA style whether it's a law or recommendations or an executive order, however that might be. My dream is that they all they're harmonized that they they are similar and they're not putting conflicting requirements on on everybody's plates. >> Yeah, that sounds reasonable.
19:50 Let's see what will come out from it. [laughter] Probably one of the most important thing that I I know about the CRA is well the act was in some way similar but the CRA is closer to an implemented measure was the fact that the the industry and the regulators were together and you saw that they managed to come to a consensus. Now the question is in terms of implementation because now it was the debate into how things are coming to force because obviously bigger companies had bigger lobbying groups or >> yes >> and
20:32 that's my only concern but something tells me that with the issues happening in the last week or two weeks people are more aware of the importance of cyber security and also it's probably worth mentioning that there is an ongoing war it's more more than that there is a cyber war happening globally and that's one of the defenses that we have to have.
20:53 So yeah, fingers crossed that things are moving in the right direction and and to touch back on your point, you mentioned digital sovereignty and this is where our perspective is that open source is the best way for anyone to manage their own destiny, to make their own choices. And you know, open source software lets you look at the code, fork the code, change the code, provide updates to the whole community.
21:21 And that's I think a much better solution than a single solution from a single vendor. And again, it allows you to pick and choose the best of breed. So you might want to use a, you know, a Chinese openweight model for your LLM, but then you might want to use a Europeanbased uh Kubernetes platform while leveraging something that was a web server written in the States or India or Africa.
21:48 It allows you to put this together and have that insight into how was this written, how does that project work? And if you don't like that, you have the ability to take your work and move it to something else. It's your choice >> just to factor that in. There is also another movement is it seems that as you mentioned Europe is very strong on the individual and his or her rights and I spoke a lot and I looked a lot into local first in the last couple of months and that's exactly the story that local first tries to push
22:27 >> that we shouldn't rely on a cloud and doesn't matter on which side of the Atlantic Ocean or Indian Ocean Wherever it's being situated, your data is your data and rather than getting something like a cryptic JSON file when a given product is sunset, the data remains with you and you can use it >> and that's important and uh hopefully this will just push forward together with the legislation in place and move in the right point because it's actually needed.
22:58 >> Agreed. >> But whether we like it or not, AI is here to stay. >> It's not going away. Yes. [laughter] Yes. Well, if we are looking at the doomers, we are the ones going away and everything will be nice and shiny for the AI. Are you an optimist or uh on the contrary or you're somewhere in the middle? I put myself as an AI realist and I use it in my own practice today.
23:24 It's a helpful assistant and it helps accelerate. I don't see it as replacing me unless Skynet is going to plug me in as a battery, you know, however that might work out. But again, I'm a realist. It's just like a fuzzer or like a voltmeter I might use to measure like current in electronics. It's a tool and it's how you use that tool is what types of outcomes you're going to get.
23:48 I hope people recognize the tra the good and bad aspects of using AI. LLMs are math, not magic. They're basically guessing what the next word will be based off of the words that came before it. That context, that's all it is. And roughly a third of the time it's wrong. And that's where ideally people will recognize where it's strong. It's amazing. It's sifting through huge volumes of data and recognizing patterns, which is something the human mind is a little weak at.
24:18 But that's where you know used judiciously and correctly it can be an amazing helper. But if you are not paying attention or you are again you don't understand the other security practices or things you need to do it's very easy to exfiltrate your personal data your company data out to the model that potentially other people could look at it. So again, I'm a realist and I try to advocate, make people aware of the benefits and the potential negative consequences of using this stuff.
24:51 >> Okay, fair enough. Given that we spoke about open source and we spoke about AI as the tool of choice for most of us these days, there is a lot of I would not call it noise in the open source community. But there are a couple of debates about AI policy. For instance, you have on one side the kernel developers which are quite strong on everything has to be human touched.
25:15 Then you have Lenus Tor which has a debate and says okay AI is another tool use it cautiously but you have to understand what's there and so on so forth. What's your perspective? How do you see the AI policy of I don't know recommended maybe >> for upstream it is still not finalized different communities have had different impacts based off of researchers using these models to find and throw vulnerability reports at them.
25:47 So we're kind of at a crossroads today where there are some people like right now the kernel has embraced the possibility of using AI as part of the code contributions but other communities that might not have access to resources or that training they will have a counter opinion. So again I think in 2027 you're never going to make AI go away. The barn door has been opened and the horses have gotten out so to speak.
26:13 But it's just a matter of how can these projects adapt their processes to explicitly state what their beliefs are. You can use AI, you cannot use AI with us, and then kind of understand the consequences of that where people might make different choices to participate in different projects because they can leverage an assistant with some or can't with others.
26:35 >> Let's hope for a brighter future. Thank you for your time and see you down the road. >> Excellent. It was very nice to meet you. Thank [music] you for the time. >> [music] [music] [music]