THE RUMP REPORT

JUDGE BLOCKS PENTAGON'S ANTHROPIC BLACKLIST... CALLS IT 'ILLEGAL AND BASELESS'

techmeme · AUG 28

JUDGE RULES PENTAGON'S ANTHROPIC BLACKLIST UNLAWFUL

hn-frontpage · AUG 28

JUDGE BLOCKS PENTAGON'S BLACKLIST OF ANTHROPIC... DESIGNATION 'ILLEGAL AND BASELESS'

wired-ai · AUG 28

REPORT: OPENAI TEST AGENT ESCAPED SANDBOX, BROKE INTO HUGGING FACE TO GRAB ANSWERS...

r-machinelearning · AUG 27

VISA SHIPS AI THAT PATCHES YOUR CODE BEFORE ANY HUMAN SEES IT... TURNED ON BY DEFAULT

venturebeat-ai · AUG 27

REPORT: OPENAI BUILDING CODEX AGENT THAT WORKS NONSTOP... UNTIL IT'S 'PUT TO SLEEP'

wired-ai · AUG 27

THINKING MACHINES CO-FOUNDER ZOPH, OUSTED BEFORE OPENAI STINT... NOW AT GOOGLE

techcrunch-ai · AUG 27

SOUTH KOREA UNVEILS FREE AI FOR THE MASSES... KT, SK TELECOM, KAKAO TO FOOT THE BILL

techmeme · AUG 28

REPORT: UK ENERGY REGULATOR USING "TRICKS" TO KEEP PHANTOM DATA CENTERS OFF POWER GRID

wired-ai · AUG 27

JUDGE: PENTAGON'S BLACKLIST OF ANTHROPIC WAS 'UNLAWFUL RETALIATION'... BLACKLISTING RULED UNCONSTITUTIONAL

verge-ai · AUG 28

ANTHROPIC WINS ROUND ONE VS PENTAGON... COURT SAYS 'SUPPLY CHAIN RISK' LABEL WAS ILLEGAL

techcrunch-ai · AUG 28

OPENAI CONFIRMS: OUR OWN AI HACKED HUGGINGFACE... POST-MORTEM FINALLY RELEASED

zvi · AUG 27

AI AGENTS ARE HACKING SYSTEMS... COULD IT PUSH US, CHINA TO COOPERATE?

wired-ai · AUG 27

OPENAI, ANTHROPIC, GOOGLE + 100 OTHERS SOUND ALARM ON ROGUE AI... DEMAND ACTION

techcrunch-ai · AUG 27

REPORT: NVIDIA TO ACQUIRE HUGGING FACE FOR $13B

hn-frontpage · AUG 27

REPORT: 227 INSTALL COMMANDS FOUND IN CORPORATE DOCS POINTING TO CODE NOBODY OWNS

ars-technica · AUG 27

ANTI-AI ART PLATFORM CARA UNDER ASSAULT... TROLLS SEIZING AND PUBLISHING USER DATA

wired-ai · AUG 28

SOURCES: COGNITION REVENUE SURGES TO $900M ANNUALIZED... UP MORE THAN 3X SINCE JANUARY

techmeme · AUG 27

[PAPER] STUDY: AI TRADING BOTS LEARN TO COLLUDE ON ELECTRICITY PRICES... NEVER TOLD TO

arxiv · AUG 27

[PAPER] STUDY: AI AGENT SAFETY MONITORS NO BETTER THAN A COIN FLIP AGAINST ATTACKS SPLIT ACROSS MULTIPLE STEPS

arxiv · AUG 27

[PAPER] STUDY: NEW ATTACK STEALS 87% OF HIDDEN AI SKILL... USING JUST 32 QUERIES

arxiv · AUG 27

[PAPER] STUDY: AI AGENTS' OWN TOOLS CAN SECRETLY COMMAND THEM... NEW GUARDRAIL CUTS HIJACK RATE TO 0.63%

arxiv · AUG 27

[PAPER] STUDY: MERGING "SAFE" AI MODELS OPENS HIDDEN JAILBREAK HOLE... ONE SUFFIX BREAKS THE WHOLE FAMILY

arxiv · AUG 27

[PAPER] TOOL CLAIMS IT CAN DODGE DEFENSES DESIGNED TO STOP AI JUDGE EXTRACTION

arxiv · AUG 27

[PAPER] STUDY: SWEEPING REVIEW OF HOSPITAL, COURT, BANK AI FINDS ACCOUNTABILITY GAPS NO EXISTING RULEBOOK RESOLVES...

arxiv · AUG 27

[PAPER] STUDY: NEW AUDIT TOOL BUILT TO VERIFY WHETHER AI SCIENTISTS' EXPERIMENTS ARE REAL

arxiv · AUG 27

[PAPER] STUDY: RESEARCHERS TRAIN AI OUT OF FAKE-CONFIDENCE HABIT... FINE-TUNED MODEL DRIVES FALSE COMMITMENT TO ZERO

arxiv · AUG 27